INTERNATIONAL SERIES IN PURE AND APPLIED PHYSICS G. P. HARNWELL, Consutrine Eprror Apvisory Eprroriat Commrrree: E, U. Condon, George R. Harrison, Elmer Hutchisson, K. K. Darrow QUANTUM MECHANICS Donated by Mrs. Yemuna Bappu to The Indian Institute of Astrophysics from the personal collection of Dr. M. K. V. BappuINTERNATIONAL SERIES IN PURE AND APPLIED PHYSICS G. P. HarnwELL, Consulting Editor Bacusr anp Goupsmit—ATOMIC ENERGY STATES Birrer—INTRODUCTION TO FERROMAGNETISM Bamours—WAVE PROPAGATION IN PERIODIC STRUCTURES Capy—PIEZOELECTRICITY Cuark—APPLIED X-RAYS Curmis—ELECTRICAL MEASUREMENTS Daver—CRYSTAL STRUCTURE AND ITS APPLICATIONS Epwarps—ANALYTIC AND VECTOR MECHANICS Hagpy anv Perrmix—THE PRINCIPLES OF OPTICS Harnwaut—ELECTRICITY AND ELECTROMAGNETISM Harwwaut np Livivcoop—EXPERIMENTAL ATOMIC PHYSICS Houstox—PRINCIPLES OF MATHEMATICAL PHYSICS Hucues anv DuBawcz—PHOTOELECTRIC PHENOMENA Hunp—HIGH-FREQUENCY MEASUREMENTS PHENOMENA IN HIGH-FREQUENCY SYSTEMS Incersout, Zopet, ap IncrRsoi—HEAT CONDUCTION Kewerr—PRINCIPLES OF QUANTUM MECHANICS Kennarp—KINETIC THEORY OF GASES Kouter—THE PHYSICS OF ELECTRON TUBES Morsz—VIBRATION AND SOUND Paviane an Goupsmm—THE STRUCTURE OF LINE SPECTRA Ricuturer anp Kewwarv—INTRODUCTION TO MODERN PHYSICS Roark ano Urey—ATOMS, MOLECULES, AND QUANTA Scutr>—QUANTUM MECHANICS Sermz—THE MODERN THEORY OF SOLIDS. Suater—INTRODUCTION TO CHEMICAL PHYSICS MICROWAVE TRANSMISSION Suarer AND Frank—ELECTROMAGNETISM INTRODUCTION TO THEORETICAL PHYSICS MECHANICS Surraz—STATIC AND DYNAMIC ELECTRICITY Srrartox—ELECTROMAGNETIC THEORY Wurrs—INTRODUCTION TO ATOMIC SPECTRA Wititams—MAGNETIC PHENOMENA Dr. Lee A. DuBridge was consulting editor of the series from 1939 to 1946.QUANTUM MECHANICS BY LEONARD I. SCHIFF Professor of Physics ‘Slanford University Seconp Ivo PRSLON NEW YORK ‘TORONTO LONDON McGRAW-LULL BOOK COMPANY, ING. 1949 uA Lib. SAINTQUANTUM MECHANICS Copyright, 1949, by the McGraw-Hill Book Company, Ine. Printed in the United States of America. All rights reserved. ‘This book, or parts thereof, may not be reproduced in any form without permission of the publishers. ‘THE MAPLE PRESS COMPANY, YORK, PA.PREFACE This volume has a threefold purpose: to explain the physical concepts of quantum mechanics, to describe the mathematical formalism, and to present illustrative examples of both the ideas and the methods. The book is intended to serve as a text at the graduate level and also as a reference book. It is assumed that the reader is reasonably familiar with atomic structure, classical mechanics, and differential equations. In addition, he should have had some contact with electromagnetic theory and, for the latter part of the book, with the special theory of relativity. The author believes that the analytical methods employed in the book will satisfy most theoretical physicists even though no attempt is made to achieve mathematical rigor. For example, there is little or no dis- cussion of the justification for the interchange of sum, derivative, and integral operations, or for the use of the 5 function. On the other hand, the physical reasons for the nature of the results obtained are investigated wherever possible. Problems are given at the end of each chapter. They are often used to illustrate or amplify points discussed in the text. Original theoretical papers are referred to throughout the book; the list is representative rather than exhaustive. Experimental results are, for the most part, quoted without reference, since the large amount of documentation required for an adequate survey seems out of place in a book on theoretical physics. Several other books on quantum mechanics jects are referred to for more detailed discussions of particular topics. ‘The scope of this volume ix best outlined if the book is divided into three parts. The first three chapters constitute an introduction to quantum mechanics, in which the physical concepts are discussed and the Schrédinger wave formalism is established. ‘The detailed treatment of the wave function (Chap. IIT) may be omitted in a first reading. The next cight chapters comprise the central part of the book. This part presents exact solutions of the wave equation for both energy-level and collision problems, the Heisenberg m: formalism and transformation theory, approximation methods, radiation theory, and some applications to atomic systems. Since the first eleven chapters correspond to a typical one-year graduate course, it seemed desirable to include a semi- classical treatment of electromagnetic radiation in the central part of the book (Chap. X) even though some of the results are obtained again in Chap. XIV. The last part of the book corresponds to a short course in what is often called advanced quantum mechanics. It consists of rela- vPREFACE cle theory and an introduction to quantized field theory and .ectrodynamics. all this book was written while the author was at the Uni- ur Pennsylvania, and he gratefully acknowledges the continued ugement of Prof. G. P. Harnwell. Hes also indebted to Drs. E. H. a..._ard and §. Pasternack for helpful criticism of the early chapters, to W. Miller and L. Spruch for their careful reading of most of the manu- script, and to Profs. F. Bloch, R. F. Christy, and W. W. Hansen for valuable comments on certain sections. It is a particular pleasure for the author to thank Prof. R. Serber for many discussions of both the conceptual and formal aspects of quantum mechanics that took place during the last eleven years. Finally, the author wishes to acknowledge his indebtedness to Prof. J. R. Oppenheimer for his introduction to several of the ideas and exam- ples that appear in the book. Indeed, the writing of this book in its present form owes much to the period from 1937 to 1940, which the author spent in association with Professor Oppenheimer. Luonanp I. Scutrr Sranvorp, CALIFORNIA Tune, 1948its icecaeecacasacas: 10. . Baperimental Background... . A. Discussion of Measurement . CONTENTS Cuaprer T THE PHYSICAL BASIS OF QUANTUM MECHANICS. Tnadequacy of classical physics. Summary of principal experiments and inferences. The Old Quantum Theory. 0 0 Bohr-Sommerfeld quantization rules, Practical difficulties. Conceptual difficulties, Quantum-mechanical viewpoint. Uncertainty and Complementarity . Uncertainty principle. Complementarity principle. Limitations on ex- periment. ussion of diffraction experiment, Analysis of diffraction experiment. Localization experiment. Wave Packets in Space and Time... 2... : Space packets. ‘Time packets. Wave formalism. Cuavrer IT THE SCIRODINGER WAVE. EQUATION Development of the Wave Byuation 0. ‘Traveling harmonic waves. Need for a wave equa sional wave equation. Hxtension to three dim forces, Interpretation of the Wave Function. foe 5G Normalization of ¢. Probability current don sity. Expectation value, Mhrenfest’s theorem, Energy Kigenfunctions ... . ion of the oun Boundary conditions for infinite potent one dimension, Dise love Diserete and continuous ¢ One-Dimensional Square Well Potential Perfectly rigid walls, Finite poten simplified solution, ‘The one-dimen- sions, Inclusion of ance of the separntion eon- Continuity conditions, Energy cigenvalues in Continuous energy eigenvalues, 1 Unree dimensions. levels. Carrer TIT EIGENFUNCTIONS AND EIGENVALUES Interpretative Postulates and Energy Kigenfunctions Dynamical variables as operators, Mxpansion in mfunctions. The vii 12 7 a 27i 12, 13. ice 15. 16. 17. 18, 19. 20. CONTENTS total energy operator. Normalization in a box. Orthonormality of energy eigenfunctions. Reality of the energy eigenvalues. Expansion in energy eigenfunctions. The closure property. Probability function and expectation value. General solution of the Schrédinger equation. Momentum Eigenfunctions 0.2... ee Form of the eigenfunctions. Box normalization. ‘The Dirac # function, A representation of the 6 function. Normalization in terms of the 6 func tion. Some properties of the 8 function, Closure. Expansion in momentum eigenfunctions. Probability function and expectation value. Motion of a Free Wave Packet in One Dimension. ‘The minimum uncertainty product. Form of the minimum packet. Momentum expansion coefficients. Change with time of a minimum packet. Classical limit. Cuaprer IV DISCRETE EIGENVALUES: ENERGY LEVELS Linear Harmonie Oscillator... Lee Asymptotic behavior. Energy levels. Zero-point energy. Parity. Her= mite polynomials, Harmonic-oscillator wave functions. Correspondence with classical theory. Oscillating wave packet. Spherically Symmetric Potentials in Three Dimensions . . Lee Separation of the wave equation. Legendre polynomials, Spherical harmonics. Parity. Angular momentum. Three-Dimensional Square Well Potential... 2... . Zero angular momentum. Interior solutions for arhitrary /. solutions for arbitrary J. Energy levels. ee eee ee . coe Reduced mass, Asymptotic behavior, Energy levels. Laguerre poly- nomials. Tydrogen-atom wave functions. Degencracy, Separation in parabolic coordinates. Energy levels. Wave functions. Craprir V CONTINUOUS EIGENVALUES: COLLISION THEORY One-Dimensional Square Potential Barrier... 2. Asymptotic behavior, Normalization. Scattering Coefficients. rr ne Scattering cross section. Relations between angles in the laboratory and center-of-mass systems. Relation between cross sections. Dependence ony. Asymptotic behavior. Normalization. Scattering by Spherically Symmetric Potentials. ©... 2... 1. oe Asymptotic behavior. Scattering cross section. Phase shifts, Caleu- lation of & Relation between signs of &: and V(r). Ramsauer-Townsend efiect. Scattering by a perfectly rigid sphere. Scattering by a square well potential. Resonance scattering. Angular distribution at low energies. Scattering by a Coulomb Field... ee Parabolic coordinates. Confluent hypergeometric function, Scattering cross section and normalization. Solution in spherical coordinates. Modi- fied Coulomb field. Classical limit for a pure Coulomb field. 48 Bt 60 69 76 80 92 96 103 14CONTENTS ix Cuarrer VI MATRIX FORMULATION OF QUANTUM MECHANICS SManedgee . 122 Matrix addition and multiplication. Null, unit, and constant matrices. Spur, determinant, and inverse of a matrix. Hermitian and unitary matrices. Transformation and diagonalization of matrices. Matrices of infinite rank. . Matrices in Quantum Mechanics... 0. oo ee - 127 ‘A unitary transformation matrix. Hamiltonian matrix, Dynamical variables 2s Hermitian matrices. Wave functions as unitary matrices. ‘The r representation. A useful identity. }. Equations of Motion in Matriz Form. 2... 6 ee ee ee 181 Time rate of change of a matrix. Classical Lagrangian and Hamiltonian equations of motion. Poisson brackets and commutator brackets. Quan- tization of a classical system. Motion of a particle in an electromagnetic field. Evaluation of commutator brackets. Velocity and acceleration of a charged particle. The Lorentz force. Constants of the motion. The parity operator. Energy reprosontation. | Anguies Momentum 140 Definition of angular momentum. Choice of a representation, Relations between matrix clements. Eigenvalues of Ms. Eigenvalues of M?; the I, matrix. Connection with the spherical harmonies. Spin angular mo- mentum, Addition of angular momenta. Wigenvalues of (Mi -+ Mz)*. Cuaprer VIL APPROXIMATION METIIODS FOR STATIONARY PROBLEMS . Stationary Perturbation Theory oboe. . 149 Nondegoneraty ease. First-order perturhation. Second-order perturba tion, Normaliztion of y. Perturbation of an oscillator, Degenerate case. Removal of degeneracy in first order, Removal of degeneracy in second order, First-order Stark effect in hydrogen. Perturbed energy levels, Occurrence of permanent clectric-dipole moments. . The Born Approximation... .. 2... Se eee 159 Perturbation approximation, Green's function. Green's funetion for a free particle, Explicit evaluation. Scattering cross section, Perturha- tion treatment of partial waves. Phase shifts. Scattering by a square well potential, Validity of the Born approximation, Seattoring by a sercened Coulomb field. . The Variation Method. Wxpeetation value of the energy. state of helium, Ek Ago 6 100 “Application to « ion energy. Variation of the parameter Z, van der Waals intornetion. Perturbation calculation. Variation cul- culation, . The WKB Approximation . . . eS Classical limit, Approximate solutions, "Asymptotic nature of the solu- tions. Solution near a turning point, Linear turning point. Connection at the turning point. Asymptotic connection formulus, Energy levels of a potential well, A quantization rule, Special boundary conditions.x CONTENTS Cuaprer VIII APPROXIMATION METHODS FOR TIME-DEPENDENT PROBLIEMs 29, Time-Dependent Perturbation Theory... . : Expansion in unperturbed eigenfunctions, First-order perturb ion, Physical interpretation, ‘Transition probability. Scattering cross section, Harmonie perturbation. Second-order perturbation. ffect of first-ord transitions. Intermediate states. eee ee . 199 Expression for the cross section. valuation of the matrix clement, Differential and total cross sections. Production of a clowl-chamber track. Formulation of the problem. valuation of the k sum. Second~ order matrix element. Discussion of the cross section. 81. Adiabatic and Sudden Approximations... ...... - pe Expansion in instantaneous energy eigenfunctions. of phases, Adiabatic approximation. Connection with perturbation theory. Dis- continuous change in H. Sudden approximation. ‘Transient disturbance, Disturbance of an oscillator. 189 Cuarrer IX IDENTICAL PARTICLES AND SPIN 32. Identical Particles . . Physical meaning of identity. Symmetric and antisynmetrie wave furne= tions. Construction from unsymmetrized functions. Distinguishaly of identical particles, ‘The exclusion principle. Connection with stat isti- cal mechanics. Collisions of identical particles. 88, Spin Angular Momentum... .. 2... + 223 Connection between spin and statistics, Spin matrices and cigenfunctionis, Collisions of identical particles. Klectron spin functions. ‘The helium atom. Spin functions for three electrons, 34, Rearrangement Collisions... . te Notation for rearrangement collisions. Use of the Born approximation. Lack of orthogonality of initial and final states, Connection with time. dependent perturbation theory, Exchange collisions of electrons with hydrogen. Differential cross section. Iixchange collisions with heliun . 216 Cuaprer X SEMICLASSICAL TREATMENT OF RADIATION 35. Absorption and Induced Emission... ... . ee Maxwell’s equations. Plane electromagnetie wavi of perturha theory. Transition probability, Interpretation in terms of absorption and emission, Electrie-dipole transitions. Forbidden transitions, 35. Spontaneous Emission... Le Classics] radiation field. Asymptotic form. Radiuted energy. Dipole radiation. Angular momentum. Dipole case. Conversion from classical to quantum theory. Planck distribution formula. Line breadth. 37. Some Applications of Radiation Theory... . = 267 Selection rules for a single particle. Polarization of emitted radia on 28 ion,39. 41. 42, 43. wh 45, 16 CONTENTS xi Conservation of angular momentum. Selection rules for many-particle systems. Cerenkov effect. Expression for the current density. Fourier analysis of the radiation field. Radiated energy. Cuarrer XI ATOMS, MOLECULES, AND ATOMIC NUCLEI Approximations in Atomic Structure... 2... 6 > 267 Central-field approximation. Periodic system of the ele1 ements. " Thomas- Fermi statistical model. Evaluation of the potential. Hartree’s self- consistent fields. Connection with the variation method. Corrections to the central-field approximation. LS coupling scheme, Selecti i coupling scheme. The Alkali Atome. oc et ee 279 Doublet separation. Doublet intensity. Hffect of a magnetic “field. Weak-field case. Strong-field case. Quadratic Zeeman effect. ee ee eee » 288 Classification of energy levels. Wave equation, ‘The hydrogen molecu! Potential-energy function. The Morse potential. Rotation and vibration of diatomic molecules. Energy levels. Effect of nuclear identity. Alomie Nuclei, 6. 0 fs WT General properties of nucl Nature of nuclear interactions. Nuclear models, Neutron-proton system. Exchange operators. Proton-proton scattering. Ciarrer XIT RELATIVISTIC WAVE EQUATIONS Schrivtinger’s Relativistic Equation... ss. . ong Free particle. Hlectromagnetie potentials, Separation of the equation, nergy levels in a Coulomb field. Dirac’s Relativistic Bquation. . . . : [ail Freo-partiele equ ele solutions. Charge and ¢ Dirae's Equation for « Central Bicld alo Spin angular moment Approsi energy. Reparation of the equation, The hy ation of cnorgy levels. Negative energy states, Cuavese XU TUE QUANTIZATION OF WAVE. FIELDS. Classical and Quantum Pied Equations to : 330 Coordinates of the field, Lagrangian equation, Vunetional derivative. Hani n equations, Quantum equa for the field, Fields with more than one component. Quantization of the Nor relativistic Schrodinger Equation... 336 Lagrangian and Hamiltonian equations, Quantum equations. “The N representation, Connection with the harmoni interpretation. Conneetion with the many-purticle Sehrddinger equation,xii CONTENTS Anticommutation relations. Equation of motion. Physical implications of anticommutation. Representation of the anticommuting a, operators. 47. Quantization of the Dirac Equation. rn Lagrangian and Hamiltonian equations. Quantum equations. ‘The N representation. Negative energy states and positrons. Anticommutation relations at different times. Commutation relation for the charge density. Cuaprern XIV QUANTUM ELECTRODYNAMICS 48. Electromagnetic Field in Vacuum Lagrangian equations. Hamiltonian equations. Quantum: equations. Commutation relations for E and H. Plane wave representation. Quan- tized field energy. Quantized field momentum. A(r,t) in the plane wave representation. Commutation relations at different times. 49. Interaction Between Electrons and the Electromagnetic Field . . ‘Lagrangian and Hamiltonian equations. Elimination of ¢. Quantization of the fields. Inclusion of static fields. Use of perturbation theory. Matrix elements of the Coulomb interaction. 50. Radiation Theory . . Formulation in terms of transition probsbilities, Matrix elements of the perturbation. Transition probability for absorption. Transition proba- bility for emission. Analysis of a diffraction experiment. Representation of the electromagnetic field. Matrix elements. Classical diffraction pattern, 348 . 361 . 873 382 . 397CHAPTER I THE PHYSICAL BASIS OF QUANTUM MECHANICS At the present stage of human knowledge, quantum mechanics can be regarded as the fundamental theory of atomic phenomena. The experimental data on which it is based are derived from physical events that lie almost entirely beyond the range of direct human perception. It is not surprising, therefore, that the theory embodies physical con- cepts that are foreign to common daily experience. These concepts did not appear in the historical development of quantum mechanics, how- ever, until a quite complete mathematical formalism had been evolved. The need for quantitative comparison with observation, which is the ultimate test of any physical theory, in this case led first to the formalism and only later to its interpretation in physical terms. It seems desirable in introducing the subject of quantum mechanics to depart from the historical order and preface the mathematical develop- ment with a discussion of the physical concepts. In this chapter we first review briefly the experimental background and the ideas of the old quantum theory, then discuss the newer physical concepts of uncertainty and complementarity, and finally lay the groundwork for the formalism that will be developed in its most familiar form in Chap. II. No attempt will be made to deduce the structure of the formalism from the funda- mental experiments; we shall try to make the theoretical development seem plausible rather than unique. The justification for the theory, then, will rest on the agreement between deductions made from it and experiments, and on the simplicity (in principle more than in practice) and consistency of the formalism. 1. EXPERIMENTAL BACKGROUND Experimental physics prior to 1900 had demonstrated the existence of a wide variety of phenomena, which for the most part were believed to be explicable in terms of what we now call classical theoretical physics. The motions of mechanical objects were successfully discussed in terms of Newton’s equations on both celestial and terrestrial scales. Appli- cation of this theory to molecular motions produced useful results in the kinetic theory of gases, and the discovery of the electron by J. J. Thom- son in 1897 consisted in showing that it behaved like a Newtonian particle. 12 QUANTUM MECHANICS [Cnar. T The wave nature of light had been strongly suggested by the diffraction experiments of Young in 1803, and was put on a firmer foundation by Maxwell’s discovery in 1864 of the connection between optical and electrical phenomena. Inadequacy of Classical Physics. The difficulties in the understand- ing of experimental results that remained at the beginning of this century were largely concerned with the development of a suitable atomic model and with the late discoveries of X rays and radioactivity. However, there were also difficulties associated with phenomena that should have been understood but actually were not: such things as the spectral dis- tribution of thermal radiation from a black body, the low-temperature specific heats of solids, and the appearance of only 5 degrees of freedom in the motion of a free diatomic molecule at ordinary temperatures. The beginning of an understanding of the second class of difficulties was made by Planck in 1900, when he was able to explain the black-body spectrum in terms of the assumed emission and absorption of electro- magnetic radiation in discrete guanta, each of which contains an amount of energy E that is equal to the frequency of the radiation »v multiplied by a universal constant h (called Planck’s constant): E=hy (a) This quantum idea was later used by Einstein in accounting for some of the experimental observations on the photoelectric effect. In this way the dual character of electromagnetic radiation became established: it sometimes behaves like a wave motion, and sometimes like a stream of corpuscular quanta. At about this time, the existence of discrete values for the measurable parameters of atomic systems (not only of electromagnetic radiation) became apparent through Einstein’s and Debye’s theories of the specific heats of solids, Ritz’s classification of spectral lines, the experiment of Franck and Hertz on the discrete energy losses of electrons on collision with atoms, and (somewhat later) the experiment of Stern and Gerlach, which showed that the component of the magnetic moment of an atom along an external magnetic field has discrete values. Summary of Principal Experiments and Inferences. The theorcticul physics of the first quarter of this century thus contained two important inferences, obtained from the experiments and their interpretations, that had not existed in 1900: the dual character of electromagnetic radiation, and the existence of discrete values for physical quantities. The relations between the principal experimental conclusions and the theorctical inferences are shown schematically in Table 1; for a more detailed dis-Sec. 2] THE PHYSICAL BASIS OF QUANTUM MECHANICS 3 cussion and a bibliography, reference should be made to a book on atomic physies.! Taste l. RELATIONS BETWEEN EXPERIMENTAL INTERPRETATIONS AND THEORETICAL InvERENCES Diffraction (Young 1808, Laue 1912).......0......cceeeeeeeeces — Black-body radiation (Planck 1900) Photoelectric effect (Einstein 1904) {Blectomagneti (Compton'ettest((1020) at quanta Combination principle (Ritz-Rydberg 1908). ; Specific heats (Einstein 1907, Debye 1912) Pissalied Ate Franck-Herts experiment (1913) [UT : aati Stern-Gerlach experiment (1922) A third theoretical inference appeared in 1924 with the suggestion by de Broglie that matter also has a dual (particle-like and wave-like) character; he assumed that the relation between the momentum p of the particle and the length of the corresponding wave is? (1.2) Up to that time all the evidence had indicated that matter was composed of discrete Newtonian particles; in particular, sharp tracks of charged particles such as clectrons and helium nuclei had been observed in expan- sion cloud chambers like that invented by C. T. R. Wilson in 1911. Shortly after this, however, Davison and Germer (1927) and G. P. Thomson (1928) independently observed the diffraction of electrons by crystals, and thus confirmed de Broglic’s principal supposition. 2. THE OLD QUANTUM THEORY What is now called the old quantum theory* was initiated by the work of Planck on black-body radiation, and carried farther by Finstein and Debye. However, it was only after Rutherford’s discovery in 1911 that an atom consists of a small, massive, positively charged nucleus sur- rounded by electrons, that the theory could be applied to a quantitative description of atoms. ¢, for example, chumyer and [2 TL Kennard, “Introduction to Modern Physies” (McGraw-Hill, New York, 1942); M. Born, “Atomic Physies” (Blackie, Glasgow, 1946). * Equation (1.2) is also valid for light quanta, as may he seen by dividing both sides of Eq, (1.1) by the velocity of light ¢; for a directed beam of light p = E/e and dA = c/v. # For a more detailed discussion than is presented in this section, see the books cited above, and I., Pauling and 15. 8. Wilson, Jr., “Introduction to Quantum Mechanies,” Chap. IT (MeGraw-Hill, New York, 1935).4 QUANTUM MECHANICS (Cuar. I Bohr-Sommerfeld Quantization Rules. The first step in this direc- tion was taken by Bohr in 1913, when he made two postulates concerning the electronic or extranuclear structure of an atom. The first of these was that an atomic system can exist in particular stationary or quantized states, each of which corresponds to a definite energy of the system. Transitions from one stationary state to another are accompanied by the gain or loss, as the case may be, of an amount of energy equal to the energy difference between the two states; the energy gained or lost appears as a quantum of electromagnetic radiation, or as internal or kinetic energy of another system. The second postulate (in agreement with that of Planck and Einstein) was that a radiation quantum has a frequency equal to its energy divided by Planck’s constant h. These two postulates by themselves provided some insight into the Ritz combination principle and the Franck-Hertz experiment. To obtain specific results for hydrogen, Bohr proposed a simple rule for the selection of those circular orbits which are to constitute stationary states: the angular momentum must be an integral multiple of h/2r. A more general quantization rule was discovered independently by W. Wilson (1915) and by Sommerfeld (1916), thus making possible the application of Bohr’s postulates to a wider variety of atomic systems. This rule is applicable to Hamiltonian systems in which the coordinates are cyclic variables, and states that the integral of each canonical momentum with respect to its coordinate over a cycle of its motion must be an integral multiple of h. The rule was applied with considerable success to the computation of the fine structure of hydrogen, the spectra of diatomic molecules, and other problems. Practical Difficulties. The old quantum theory encountered practical difficulties in several different respects. It could not be applied to aperiodic systems, it provided only a qualitative and incomplete treat- ment ofthe intensities of spectral lines, and it did not give a satisfactory account of the dispersion of light. Moreover, improvements in experi- mental techniques soon showed that there were problems, such as the rotational spectra of some diatomic molecules, to which the theory gave unambiguous but incorrect answers. The correspondence principle was introduced by Bohr in 1923 in an effort to make use of the classical theory as a limiting case to infer some properties of atomic systems, especially the intensities of spectral lines. Although much was achieved in this way, it was clear in the early 1920’s that the quantum theory as it then existed was unsatisfactory. Conceptual Difficulties. Quite apart from the practical difficulties outlined above, the old quantum theory failed to give a conceptually satisfactory account of the fundamental phenomena. It was difficult toSue. 2] THE PHYSICAL BASIS OF QUANTUM MECHANICS 5 understand why the electrostatic interaction between a hydrogen nucleus and an electron should be effective when the ability of the accelerated electron to emit electromagnetic radiation disappeared in a stationary state. The mechanism of emission and absorption of radiation in transi- tions between stationary states was obscure. The quantization rules were arbitrary even when they were most effective. And the assumption of a dual character for light (particle-like on emission and absorption and wave-like in transit) seemed to be self-contradictory. In order to illustrate the conceptual difficulties and the way in which they are dealt with by the new quantum mechanics, we consider in some detail a simple diffraction experiment, which is illustrated schematically in Fig. 1. A light source S illuminates a diaphragm A in which two slits la lB Fig. 1. A diffraction experiment in which light. from 8 passes through the two slits in A to form a diffraction pattern at B. are cut. A diffraction pattern appears at a photosensitive screen B, and the ejected photoelectrons are most numerous at the diffraction peaks. Here we have the radiation behaving as a wave during its passage from source through slits to screen, but behaving as a stream of light quanta or photons when it ejects electrons from B. We now know that a similar experiment could be set up with matter instead of radiation. The diffraction pattern of electrons scattered from a crystal (analogous to the slits in A) may be detected as a distribution of electron tracks in a Wilson cloud chamber (analogous to the screen B), so that the wave and particle aspects of matter appear in the same experiment. In the situation illustrated in Fig. 1, we might at first suppose that the diffraction pattern is due to an interference between different photons passing through the two slits, thus explaining the observations entirely in terms of the particle picture. That this is not a sufficient explanation may be shown by decreasing the intensity of the light until an average of only one photon at a time is in transit between source and sereen. The diffraction pattern still appears as the distribution of the large number of photons accumulated over a sufficiently long time. ‘Thus we must con-6 QUANTUM MECHANICS [Cuav. 1 ‘clude thst diffraction is statistical property of a single photon, and does not involve an interaction between photons. From the point of view of the particle picture, we may then ask how it is that a stream of inde- pendent photons, each of which presumably can go through only one of ‘the slita, ean produce a diffraction pattern that appears only when both slits are open. Or to put the question in another way, how can the presence of a slit through which a photon does not go prevent that photon from reaching a part of the soreen it would be likely to resch if that slit were closed? Quantum-mechanical Viewpoint. In this question is implicit the assumption that the photon actually does go through a particular one of the twoslits, ‘This assumption is natural from the point of view of the classical theory or the old quantum theory since these theories regard a photon or other particle as having a definite and determinable position at each instant of time. ‘The quantum mechanics, however, diseards this assumption, and asserts instead that the position of a photon has meaning only when the experiment includes a position determination. Moreover, ‘this part of the experiment will affect the remainder of the experiment ‘and cannot be considered separately. ‘Thus from the point of view of quantum mechanios, the question asked in the Inst paragraph is without meaning, since it assumes that the photon gocs through a particular one of the two slits (thus making it possible to close the other slit) when there is no provision in the experiment for determining through which slit the photon actually goes. ‘The quantum mechanics resolves the sittation by telling us that the diffraction pattern is destroyed if u sufficiently careful attempt is mace to determine through which slit each photon pastes (see Sec. 4). We must then be prepared to forego the customary mental picture of a photon (or an electron) as a classical particle that has ut euch instant of time & position that ean be determined without damage to diffraction patterns of the type discussed here. Thus classical enusality, which requires that the motion of a particle at any time bo uniquely determin- able from its motion st an earlier time, must also bo abandoned. ‘The new theory that is forced upon us in this way is s0 suecossful in other respects as well that, at the present state of knowledge, we must regard such classically incomplete descriptions us a fundamental properly of nature, 8. UNCERTAINTY AND COMPLEMENTARITY Before presenting a more quantitative dixeussion of tho diffraction experiment outlined in See. 2, we consider two principles that express in qualitative terms the physical content of tho theory of quantumSec.8] THE PHYSICAL BASIS OF QUANTUM MECHANICS 7 mechanics. We restrict ourselves here to a discussion of the and give arguments for their validity in Seo. 4 Uncertainty Principle. ‘The first of these is the uncerlainty principle, developed by Heisenberg! in 1927. According to this principle, it is impossible to specify precisely and simultaneously the values of both ‘members of particular pairs of physical variables that describe the behav- ior of an atoinic system, ‘The members of these pairs of variables are ‘canonically conjugate to each other in the Hamiltonian sense: a rectangu- lar coordinate z of a particle and the corresponding component of momen- tum px a component J, of angular momentum of a particle and ita angular position ¢ in the perpendicular (zy) plane, the energy E of a par- ticle and the time ¢ at which itis measured, ete, Put more quantitatively, the uncertainty principle states that the ordor of magnitude of the product of the uncertainties in the knowledge of the two variables must be at least Planck’s constant h divided by 2r (A = h/2r = 1.054 X 10 ceng-seconi), xo that meaning, Ar: Ape> h @n Ag+ AS, G2) At AB Sh 3) Tho relation (8.1) means thwt a component of the momentum of a particle cannot. bo precisely specified without our loss of all knowledge of the corresponding component of its position at that time, that a particle cannot he precisely localizod in a particular direction without our loss of all knowledge of its momentum component in that direction, and that in intermodliate eases the product of the uncertainties of the simultaneously measurable valuox of cortosponding position and momentum component is at leant of the ordor of magnitude of A. Similarly, Faq. (3.2) means, for examplo, that tho procise measurement of the angular position of a. particle in an orbit carries with it the lors at that time of all knowledge of the component of angular momentum perpendicular to the plane of the orbit, Wquution (8.8) means that an energy determination that has an acenmey AB must occupy at least a time interval At ~ A/a; thus if a xystem maintains a particular state of motion not Touger than atime A’, the energy of the xystem in that sate is uncertain by at Teast the amount Aff ~ h/Al, since At isthe Iongest. Lime interval available for the energy determination. ‘The smallness of fi makes the uncertainty principle of interent primarily in onncetion with aystems of atomic sie, W. Hoiwenbong, Zea f. hyn, 48, 172 (1927). FTL, Binge, Res. Mad. Phye, 48, 288 (1041),8 QUANTUM MECHANICS (Cxar.1 ‘As we shall see in Sec, 12, the uncertainty principle may be obtained directly from the mathematical formulation of the theory, and this is actually the way in which it vas first obtained by Heisenberg, Complementarity Principle. In order to understand the implications of the uncertainty principle in more physical terms, Bohr‘ introduced the complementarity principle in 1928. This principle states that atomic phenomens cannot be described with the completeness demanded by classical dynamics; some of the elements that complement each other to make up a complete classical description are actually mutuslly exclusive, and these complementary elements are all necessary for the description of various aspects of the phenomena. From the point of view of the experi- ‘enter, the complementarity principle asserts that the physical apparatus available to him has such properties thet more precise measurements ‘than those indicated by the uncertainty principle cannot be made. This is not to be regarded as a deficiency of the experimenter or of his techniques. It is rather a law of nature that, whenever an attempt is ‘made to measure precisely one of the pair of canonical variables, the other is changed by an amount that eannot be too closely ealculated without interfering with the primary attempt. This is fundamentally different from the classical situation, in which » measurement also disturbs the system that is under observation, but the amount of the disturbance can be caleulated and taken into account. ‘Thus the complementarity principle typifies the fundamental limitations on the classical concept that the behavior of atomic systems can be described independently of the means by which they are observed. Limitations on Experiment. In the atomic field, we must choose between various experimental arrangements, each designed to moasure the two members of a pair of canonical variables with different degrees of precision that are compatible with the uncertainty relations. In particular, there are two extreme arrangements, each of which measures one member of the pair with great precisiou. According to classical theory, these extreme experimental arrangements complement each other; the results of both may be obtained nt once and are necessary to supply complete classical description of the system. In quantum mechanics, however, the extreme complementary experiments are mutually exelusive ‘nd cannot be performed togethe tis in this sense that the classical concept of causality disappears in the atomic field, There is causality in so far as the quantum laws that describe the behavior of atoms are perfectly definite; there is not, bow- ever, a causal relationship between successive configurations of an 1N. Bohs, Nature, 121, 880 (1028); “Atomic Theory and the Description of ‘Natuse" especialy Part 11 (Cambridge, London, 1984); Phe, Re, 48, 608 (1035).Sec. 4] THB PHYSICAL BASIS OF QUANTUM MKCHANICS 9 atomic system when we attempt to describe these configurations in classical terms, 4 DISCUSSION OF MEASUREMENT In this section we consider two fairly typical measurement experi- ‘ments from the point of view of the new quantum mechanics, ‘The first is the diffraction experiment that was brought: up in Sec. 2, and the second is an experiment designed to determine the position and momen- tum of a particle by an optical method. Analysis of Diffraction Experiment. We fimt analyze the diffraction experiment of Seo, 2 from the point of view of the complementarity principle, assuming that the uncertainty principle is valid. ‘Two con~ trasting arrangements, which would complement euch other classically, ne Ib Vio. 2. The exvorinealal arunuomont of Ma. 1, woiifel by tho addition of photon indloctom 2 aro considered here. One of these is illustrated in Tig. 1. Since it is assumed that the distance from A to is large compared to the distance Ddetween the two slits, and thin in tum is large compared to the wave length of the light, the distvibution of intensity in the diffraction pattern determines to good approximation the :mgulur distribution of the photons leaving the slits in A, and hence determines. the distribution of the y components of momentum of the photons beyond A, ‘The xocond arrangement, shown in Fig, 2, determines Uhrough which of the two slits euch photon passes, and lence provides information on the y coordinates of the photons. In the second arrangement euch photon wiislens itself as it passe through a slit by bouncing off one of x number of indicator C placed close to A, and giving up to it a y component of momentum that may be uncertain by the amount Apy. If we do not want the resultant diffraction pattern of many such photons to be destroyed by these events, the ‘uncertainty in py for a particular photon proxduecd by its encounter with an indicator must be substantially smaller than would be required to10 QUANTUM MECHANICS (Cane. ‘throw the photon from @ maximum of the diffraction patter at B into a neighboring minimum. With a photon of momentum p., this requires ‘that APs < Ope 4 For the simple ease in which >> a>> 2, the angle 0 is known exper mentally (or from the theory of wave optics) to be given by i an in terms of the optical wave length » and the distance a between the slits. ‘At the same time, we have not learned through which slit this photon passed unless the uncertainty Ay of the y position of the indicator that recoiled is less than half the distance between slits, Ay n ficlda* We shall consider here 4.3 sa” exper eme fr th ry with which the x components jmatiation' ofa tnmlde Phy position andl momentum voetors of a he sitar material partiele can be determined at the feos me by obworving the partile through rere 8 rathor idealized mietmonpe by means of weattorod light. “The bent rsalving posver of the lens J shown in Fig. 3 is known (again yy cithoe experimentally or from Ube theory of wave opties) to provide un oa rw she 4.0) in a position determination, whore 2 is Uho wave length of the ruation that enter the lens, and € is the half angle subtended st the particle + Chaplee VExbow tho equivalenes of the wave theory of Chap. Hand the genom ‘qountun-meelisnien theory of partie, a fara mar i concerned 4 Clipler NTV shows how the theory of the elegtromagnetie fold ean be modiiod to inl quit eft, ee, for example, W. Telenberg, “The Physical Prineiples of the Quanta. Theory," Chaps. Hf, HHL (University of Chicago Prose, Chiewgn, 1980),: quanto sencHaNro8 Iman, y pin which omly One of Jens. For simplicity, we consider the case 18 wause of The fini he Het gua @vared oto nao. Pane of the Finite aperture of the ens, the precite dirt nae tne into the les isnot known, ‘Then since a. (4 pena iy piri ‘tum of the photon after it is scattered is Keren fli ad component of its momentum is approximately (iN) sit & ‘Tho > components of the momenta of the photon aul the paar Ge can be ascuraely known before the satering th ltr wines Chere ie no need then to know the z components of their Pratoms ick eee Position measurement refers to the displacement of the purticle, wile Fespet tothe mierowope, there is no reason why the toa matt vem, oof the system (particle, photon, and microscope) nel be alter deinnag tae seateng, The the uncertainty pin the om ponent of Ie meenoni ttum of tho particle after the scattoring is equal Uy tho cormespenvelinage uncertainty for the photon. rlicle ap. ~bsine. Ch) If we combine Eq. (46) with Bag, (4.7), we seo thot, just nfter the seat Gera “eee 1.8) Az Apewh is the best that we can do for the particle. Thus a realistic acconntat Gna of the properties of the radiation gives a result in agreement with Ue uncertainty relation (8.1) for the particle. ‘This experiment may also be considered from the point of view Of ary arrangenents i Te n: aulliciently ssreagedL just the complementarity prineiple, The eomplemet in the choice of wave length of the observed rac > permits an accurate determination of the position af (lie part after the scattering process, and lange } of its: mon 5. WAVE PACKETS IN SPACE AND TIME ‘The relation (4.4) between momentum and wave length, whieh is: Known experimentally to be valid for both photon and yulicles, stags gests that it might be possible to use concentruted Inuchix uf waves Ler describe localized particles of matter and quanta of rulintion, ‘Tu fix exter. ideas, we shall consider a wave amplitude or wave function (ai slepwncis evra the space coordinates zy,¢.and the time This quantity y ix amimeed ar have three basic propertis. First, it ean interfore wilh well, ws teats ite LER Account for the results of diffraction experiments. Secuml, it. ist lange in magnitude where the particle or photon is likely tole anal arqerull cuewhere, And third Y will be regarded as doserbing tie hehact safe Single particle or photon, not the statistical distribution uf 1 nanien «sfSec. 6] THE PHYSICAL BASIS OF QUANTUM MECHANICS 1B such quanta, This last is an essentisl requirement in view of the conelu- sion of Sec. 2 that a single quantum of matter or radiation interferes with itself rather than with other quanta. Tn this section we shall con- fine ourselves to a qualitative discussion of the one-dimensional case, in which the wave function y depends only on z and ¢, and leave the quan- titative development for Chap. TI. Space Packets. A typical form for a concentrated bunch of waves, which we shall eall a wave packet, is shown in Fig. 4a, where Y(z,0) is plotted against x for n particular time ¢ The average wave length fourier oe (a) eu, Plotn of tye! wave packet #6 nnd ite Fourar transfor. Ao and tho approximate extension Az of the packet are indicated in the diageam, The Fourier integral analysis! uf y with respect to 2 is now of interest sineo it shows how y may be built up out of continuous harmonic waves of various lengths. This is indicated in Fig. 4, in which the Fourier transform of ¥ is plotted schomatically against the propagation number k= I). Tt eun be shown by standurd mathematical methods that ae oy where Ak is the approximate xpread in propagation number associated with the packet. Tf now wo correlate wave length and momentum us in iq. (4.4), we see that the spread Ak corresponds to a momentum spread t= a(8) = fara teat S Combination of Eq. (5.1) with Eq. (5.2) gives de dp zh (5.3) * Seq, for example, TA. Pipes, “Applied Mathom "Chap. 111 (MeGenw-Hill, New York, 1948). for Eginows and Physi“ QUANTUM MECHANICS (Cuan. 1 which agrees with the uncertainty relation (8.1). ‘Thus the uncertainty prineiple for position and momentum of a quantum of matter or radiation follows directly from the wave-packet description and Eq, (44). ‘Time Packets. In anslogous fashion, we may examine the depend- ence of y on the time ¢ for a point « that is typical of the packet, and obtain a time Fourier transform that shows how y can be built up out of continuous hermonic waves of various frequencies ». In this case the relation between the spread in time of y and the spread in frequency of the time Fourier transform of y is sete 64) Equation (5.4) may be brought into correspondence with the uncer tainty principle by associating the energy H of a quantum with the fre- quency of the wave that represents it in a manner similar to the association of momentum with wave length given by Eq. (44). We shall make this connection through Eq. (1.1): E= ” 6.5) which may be inferred in the ease of photons from the experimental discussion of See. 1. Combination of Iq. (6.4) with Eq. (6.5) then gives the uncertainty relation (8.8). ‘The assumption that Eq. (5.5) is valid for matter as well as for radia tion may be made plausible by computing the group velocity? of a wave packet that represents a nonrelativistie particle of masé m, kinetic onergy E, and momentum p for which 2 and » are given by Kgs. (4.4) and (5.5), respectively. Tho group velocity, which is tho velocity of the center of the packet, is equal to dy _ dif _d(p*/2m) _ p a IM) "tp ap in Ga) in agreement with the classical expression for the velocity. This shows that with Eq. (6.5), the wave-packet description of the motion of Particle agrees with the classical description when the circumstances are such that we can ignore the size and internal structure of the pucket Wave Formalism. We see then that quanta of matter or racliation ccan be represented in agreement with the uncertainty prineiple by wave packets that may be superposed to produce interforenco and whose magnitudes give a measure of likelihood of location, provided that use is made of the experimentally inferred relations (4.4) and (5.5). 11 ix thon possible to set up a quantitative formalism based on the mathematical 2M, Bor, op. et, pp. 85, 26,Sno. 5] THE PHYSICAL BASTS OF QUANTUM MBCHANICS 16 analysis of wave motion. This will be done for matter in Chap. II, using the physical principles outlined in this chapter as a guide, and requiring always that the result of any calculation reduce to the result of the cor- responding claasical calculation in the appropriate limit. ‘This require- ment js a way of stating Bohr’s correspondence principle, which was mentioned in Seo. 2. At the present time, when a reasonably complete quantum theory exists, the correspondence principle is mainly of interest in assuring that the formalism has the proper classical limit, although it may also be of heuristie value in performing new ealculations or extending the boundaries of the theory. Tk might at first be thought that the exclusive use of a wave formalism for the description of matter in the next four chapters would confict, with the observed particle-wave duality discussed in Sec. 1 and hence disagree with the complementarity principle, ‘This is not the case, how- ever, for the formalism actually provides an understanding of all of the ‘measurable properties of matter, ineluding, for example, the production of Particle tracks in a cloud chamber, ‘Thus it will be shown in See. 30 that, if a single material particle is represented by a wave function of definite momentum and hence completely undetermined position, the probability of ionization of two or more molecules of the oloud-chamber gas is nogligibly small unless the molecules lie very nearly along a line parallel to the momentum veetor. Tt must be emphasized that these remarks are true only if a wave funetion of the type described in this scetion is always interpreted as represonting just one particle of matter, and not the statistical distribu- tion of a number of particles, When we wish to deseribe more than one particle, we must make tse of a wave function that depends on the eoordi- nates of ll of them. The analogous analy ical deseription of light quanta, which will bo tuken up quantitatively in Chap. XIV, makes use of a somewhat different. approuch. This is mainly beeause photons (unlike partielos ws we consider thom hero) ean be omitted or absorbed through interaction with matter, so that their number is not fixed. ‘Thus.a photon ‘wave function would have to depend on a variuble number of parameters, ‘and it is desirable to avoid such a situation, Problems 1. Give « brief diseription of each of the experinenta refer to in Table 1, togothor with thei interpretations 2, Describe briely the Daviseon-Crmor and ‘Thomson experiments, and the une ‘of the Wilsor cloud chamber for the observation of purtiele tencks. 8. A bra of silver atoms for a Stora-Cierlach experiment is produced by oollinat= Jig atoms that vaporize from lvoe bein w furnace at 1200°C. TE tho hewn travel 1 motor, uso the uncertainty relation to find the order of magnitude of the mallet pot that an be obtainod at the detestor.16 QUANTUM MECHANICS [Cnav. 1 4. Show that ia component of angular momentum ofthe electron of « hydrogen ‘atom in known to be 28 within 8 per cont error, its angular orbital position in the plane perpendicular to that component eannot be specified at al '. Use Eq, (82) to show qualitatively thet if one component of angular momen- ‘tum ofa partil ia precisely known, the two perpendicular angular-momentum com- ponents cannot be known seourntaly af the same time. (8. A Lounce rife bullet takee 0.6 econd to reach ita target. Regueding the bullet ‘as a mass point, and nogleoting effects of sir resistance and earth motion, find the order of magnitude of the spread of euecestive shots at the target under optimum conditions of aiming and fring. ', A perfectly cutio ping-pong ball is dropped in vacuum from a height equal to ‘ten time ita radius onto perfectly laatic Sxed sphere of the same radius, Neglecting cffesta due to earth motion, estimate the Inrgest number of bounces against the fixed sphere thatthe ball canbe expected to make under optimum conditions of release. & A beam of monomnergetic electrons is used to excite a particular lovel of an stom ina Franck-Herts experiment, If this eval is of short duration, owing to radia- ton back to the ground slate, show that the inelastically scattered electrons that Ihave lst energy to produce the exaited lovel will not all be expooted to have the sume final energy, If the excited level lata about 10-! second, what isthe order of maga ‘de of the elotron energy spread, mensured in electron-volts? 9, Disouss aay connections you ean think of that exist batween the throo toe ‘tainty relations (2.1), (22), and (2.3).CHAPTER II THE SCHRODINGER WAVE EQUATION is and the next several chapters are concerned with the non- in a force field that can be represented by ‘potential energy. A quantitative description of the motion in terms of ‘8 differential equation, the Schradinger wave equation, is developed in this, chapter, and applied’ to a simple one-dimensional problem, Various assumptions have to be made as regards the structure of the wave equa- tion, the boundary snd continuity conditions on its solutior physical meaning of these solutions. ‘These assumptions are degree of plausibility in this chapter and the next by relating them to experimental results, mainly those that deal with the diffraction of ‘material particles and with the existence of a classical limit to the quan- tum mechanics. However, no attempt is made to derive the formalism juely from 1 consideration of the experiments. ‘The definitive test of the theory: must, of course, be its internal consistency and the suecess with which deductions from it ageee with particular experimental meas- "urements; some examples will be worked out in Chaps. IV und V. 6. DEVELOPMENT OF THE WAVE EQUATION ‘The form of the Schrodinger wave equation is obtnined in this section by generalizing the propertios of the wave amplitude of Soc. 5. ‘The remainder of this chapter presents w discussion of some of the properticx of the equation and its solu Traveling Harmonic Waves. ‘The first task is to develop in more ‘quantitative fashion the properties of the one-fimensional wave function ¥G0) that wax divewssed qualitatively in See. 6. Tk wax shown there that for u eontinuons traveling harmonic wave, the wave length and momen= ‘tum aro relatod by Faq, (4.4) and the energy and froqueney by Ma. (5.8) Wo rewrite these two equations in terms of the universal constant Ae h/2e: 1) (6.2) A wave function ¥(¢,f) that represents a particle of completely unde- termined position travoling in the positive x direction with precisely 17 B= he,18 QUANTUM MECHANICS [onan known momentum p and kinetic energy H, would then be expected t4) have one of the forms cos (fe ul), sin (a ~ wi), ot", ‘some linear combination of them. This follows from diffraction experi ments like thaw of Davison and Germer and of ‘Thomsen (soe Nee. 1), and from the requirement that » wave packot of approximately the propagation number F and engular frequency w has & group velocity cert ‘to that ofa clasial free particle of momentum p aul onengy [seer Wey, (5.6)]. Need for @ War problem of « continuous harmonic wave, it is very desirable equation of which both the harmonic waves and more complicated waren are solutions. An example from a more familiar field of physies shoul help to clatify this point, In the case of three-dimensionsal sound vw ina gas, itis possible to obtain a solution of the problem of cof sound by a rigid sphere by superposing plane harmon in various directions. But it is far simpler to solve the differ equation for sound waves directly in spherical polar coordinates. If the ‘temperature of the gas changes from point to point, no progress eur bo ‘made in the general case without such a differential cqnation. ‘The ee rect underlying equation for sound waves ean be founul from direct eon sideration of the mechanical properties of the gus. While this ix not the ease with the equation of which the wave functions of Soo. 5 are xealti— tions, itis no less imperative to find the form of the o« ‘The need for this becomes more evident when the wave functiem ix ts deseril motion of a particle under the influence of extemal fures ‘turns out to be analogous to the propagution of xin waves inhomogeneous gas, We shall, therefore, find! an oqation for y, und ‘having found it, shall regard it a8 a more fundamental atlribule of the ‘wave functions than the harmonie forms (6.3). ‘The equation must have two basic properties, Wirt, iL linear, in order that solutions of it can be superposed lo proce in ference effects (in the three-dimensional ease) and Lo permit (he struction of wave packets. And second, the coefficients of {he exial ion ‘must involve only constants such as f and the mass smu charge of (he Particle, and not the parameters of a particular Kcinel af mation of the Particle (momentum, ‘energy, propagation number, aud frevuciey) ‘The reason for the latter requirement is that we shall Want ta lenve open the possibility of superposing solutions that belong to different vahws of these parameters; this means that such a more general weve Tune ‘eannot be a solution of an equation that invalve tine ttn (6.8) Brunton, In ode ogo oyon he ory insolent able to have aan Parameters in itsSte. 6) THE SCHRODINGER WAVE BQUATION 19 structure, Since differential equations are the easiest to handle, it is worth while to try this type first, and it turns out that the requirements can be met by a differential equation. With all these considerations in mind, we look frst at the most familiar ‘one-dimensional wave equation, that which describes the motion of ‘transverse waves on astring or plane sound waves ina gas: where 7 is the square of the wave velocity. Subs (6.3) into Eq. (6.4) shows that each of the four harmonie solutions, snd hence any linear combination of them, satisfies this differential equation, if and only if we put 6.5) where m is the mass of the particle that is to be described by Eq, (6.4) Beoause of the structure of Eq. (6.5) it is apparent that the coefficient y that appoars in Hq, ((14) involves the parameters of the motion (# or p); wo therefore discard this differential equation, The One-dimensional Wave Equation. In looking further for a suit- able equation, it ix helpful to noto that differentiation with respect to 2 ‘of wave functions like those of (6.3) hus the general effect of multipliea~ tion of the function by & (and sometimes also interchanging sine and cosine), while differentiation with respect to t has the general effect of multiplication by «, ‘Then the relation # = p!/2m, which is equivalent to the relation « = Aé?/2m, suggests that the differential equation for which we are looking contains a first derivative with respect to ¢ and second derivative with respect to 2. (68) Substitution shows that the first two of the wave functions (6.3) are not. if Tq, (0.8), but that either of the lust two may be (but not both In particular, if we choose @n then the third of the wave functions (6.3) satisfies Hq. (8.6). Moreover, the value of y given by Hq. (6.7) involves only the constants & and m. We are thus led to the one-dimensional form of the Sehrddinger! wave UM, See Ano Physik, 79, 261, 489 (1926); 81, 100 (1925). ‘The prosent treetment is somewinet different from that originally given’ by SchrOdinge.20 QUANTUM MECHANICS [emay. 12 equation for a free particle of mass m, which from Eqs. (6.6) and (6.7) may be written e aX-- ER (68) ‘The particular form in which Eq. (6.8) is written is significant in so far 1s its harmonic solution, the third of the wave functions (8.3), makes the left side Ey and the right side (p*/2m)y. ‘The fact that the solution =») is complex is not in itself a defect of the formalism. We shall have to be certain that all predicted results of posible physical observa- tions are expressible in terms of real numbers, and this will supply condition on the detailed interpretation of y. Extension to Three Dimensions. ‘The foregoing one-dimensional treatment is readily extended to three dimensions. It is natural to rewrite Eq. (6.1) as pot, k= = (69) where k is called the propagation vector. Similarly, the third of the wave functions (6.3) becomes exp {i(k — af] (6.10) where r is the position vector for the particle, Then by an obvious extension of the argument that led up to Eq. (6.8), it is seen that the ‘three-dimensional Schrodinger equation for a free particle that is repro- sented by the wave function (r,t) is vl nee Be (1) A comparison of Eqs. (6.9), (6.10), (6.11), and the classical energy equation n-e 6.12) suggests that, at least for a free particle, the energy and momentum can be represented by differential operators that act on the wave function y. Emind, p-—ihgrad (6.13) ‘The development of Secs. 7, 8, 10, and 11 will show that these are also valid representations when the particle is not free. Inclusion of Forces. ‘The next problem is to extend the free-particle ‘wave equation (6.11) s0 that it includes the effects of external forees that, may act on the particle, We shall assume for the present that theseSee. 7) THE SCHRODINGER WAVE EQUATION 21 forces are of such a nature (electrostatic, gravitational, possibly nuclear) that they can be combined into a single force F that is derivable from a potential energy V. Fort) = — grad Vir) (6.14) ‘Just ns the classical relation between energy and momentum is used above to infer the structure of Eq. (6.11), 60 it is desirable now to start from the corresponding classical relation that includes external forces. This is simply expressed in terms of the potential energy BoE + vie (6.18) ‘where H is now the total energy, and the first and second terms on the right side of Eq. (6.15) are the kinetic and potential energies of the particle, respectively. Since V docs not depend on p or B, Eqs. (6.16) and (6.18) suggest that Bq. (6.11) be gencralized into aw. — Boyt vise (6.16) ‘This is the Schrédingor wave equation that describes the motion of a particle of mass m in a force ficld given by Eq. (0.14). While the intro- duction of Hq. (6.16) cunnot claim as high a degree of plausibility as the derivation of the free-particle equation (6.11), the further discussion of the next section should make it more convincing. It is, of course, the agreement of solutions of Eq. (6.16) with experiment in particular cases that eventually demonstrates the validity and usefulness of this wave equation. 1. INTERPRETATION OF THE WAVE FUNCTION ‘The wave function (4), which is a solution of the wave equation (6.16), is now assumed to provide a quantum-mechanieally complete description of the behavior of w particle of mass m with the potential energy V(r), and hence is analogous to the clasical trajectory 1(). ‘Thus far, the only interpretativo guide available to us is that the wave funetion be large where the partiele is likely to be and small elsewhere. This has to be supplemented with more detailed statements thut enable us to get out of y the maximum amount of information permitted by nature, as was discussed in See. 8. As with the structure of the wave equation, the correctness of our interpretation of the wave function must in the end be judged by logical consisteney and appeal to experimental results,22 QUANTUM MECHANICS (Cran. 11 Statistical Interpretation. ‘The phrase ‘likely to be” in the preceding paragraph, together with the discussion of Sec. 3, indicates the need for interpreting yin statistical terms, We can imagine « very large number of identical, independent, nonoverlapping regions of space, each large enough to contain all the physically interesting features of the motion, in each of which the behavior of a particle with the potential energy V(r) is described by the same wave function v(t); in cach cuse r is, referred to the origin of the particular region, We then make the assump- tion that the numerical result of the measurement at a particular time ¢ (in so far as the time at which the measurement is made ean be specified) of any physically meaningful quantity, such as position, momentum, ‘or energy, will in general not be the same for all the regions. Rather, there will be a distribution of these numbers that can be described by probability function. For example, we have seen in Seo. 5 that the result of a position deter- rmination is to be regarded as uncertain by an amount of the order of the Linear dimensions of the wave function. It is natural therefore to regard ¥ a8 a measure of the probability of finding a particle at a particular position with respect to the origin of its region, However, a probability. must be real and nonnegative, whereas ¥ is complex in’ gencral. We therefore assume that the product of y and its complex conjugute ¥ is the position probability density. PE) = Wed? a) ‘This means that P(e,dziyde is to be the probability of finding a particle in its volume element dadyde about its point r at the timo f, when a urge number of precise position measurements are made on’ independent particles each of which is described by the one-particle wave function ve). ‘Normalization of y. Tho probability of finding the particle xome- ‘where in the region must be unity, so that Eq. (7.1) implies that the wave funetion is normalized: Siees0)Pdr = 1 2) where the integral extends over the entire region; here dr is the volume clement dadydz. If ¥ is a wave packet of the type discussed in See. 5, the integral in Eq. (7.2) converges, and the numerical cocfficient of ¥ may bbe adjusted so that the integral is unity; such normalization docs not of course change the fact that ¥ is a solution of Eq. (6.16), which is homo- geneous in y. There are, however, wave functions like that given in Eq, (6.10) for which the integral in Eb. (7.2) does not converge if taken over an infinite volume, Such wave functions require special consideraSrc. 7] THE SCHRODINGER WAVE KQUATION 23 tion, and will be discussed further in Secs. 10 and 11. For the present, wwe may think of the region of space in which such a wave function is defined as being arbitrarily large, but finite; then the integral in Eq. (7.2) is over the finite volume of this region and converges, so that normaliza- tion is always posible, ‘The cocfficient of ¥ that normalizes it must be independent of the time in order that ¥ may satinfy the wave equation (6.16). ‘Thus if Bq, (7.2) is satisfied ut-one instant of time, the interpretation of |¥]? as m position probability density requires that the normalization integral be dependent of the time. ‘That this is uetually the ease may be shown by ‘computing the time derivative of the integral of P over any fixed volume V. a f,reoa |, (03+ a)a = B [0 - wna =P [iv wera ~ eras vie Bf wemay ~ reap yea. Here substitution has been mado for a9/at from Eq. (6.16), und for af /at from the complex conjugate of Eq. (6.16). ‘The lust integral is obtained by partinl integration (use of Croen’s theorem), where A is the bounding surfaco of the region of integration and { }q denotes the component of the voctor in brackets in the direction of the outward normal to the xt face clement dA. We define a vector S(ey), Scot) ~ gh Werad y — (grad Dv] @: in terms of which é if PQ@bar = = fpaw Sir = - if Swi ay Tn the case of a wave packet, for which ¥ vanishes at. great distunees and the normalization integral converges, the surface integral is evidently zero when Vis the ontire space. Kor a wave funetion of the type given in Eq, (6.10), ¥ ean be defined in a finite region V xo that it vanishes or has a periodic structure along he bounding kurfices (see See. 10), Tn all these cases, it can be shown without difficulty hat the surface integral-int 1 1s conveniont to adopt an order of factora meh that J process ¥ [ace the die eussin of Ka. (7.7) below24 QUANTUM MECHANICS [Cuan IE Eq. (7.4) is nero, 60 that the normalisation intogral in Eq. (7.2) in con stant in time. ‘Probability Current Density. ‘The derivation of Iq. (7.4) alter shows that the differential relation ae, at is valid. ‘This has the familiar form asiociated with the consoT-Wation of flow of a fluid of density P and current density S, in which there, svn sources or sinks. Tt is thus reasonable to interpret S(r,t) given. Voy Ua. (7.8) 28 & probability current density. This interpretation malcess mor plausible the identification of —iA grad with the momentum in Ma. (0.13), even when a force is present. For then (A/im) grad is the ~elocity operator, and it is apparent that 8(6i) = real part of (¥-4; grad ») ‘While this interpretation of S is suggestive, it must be reulizecI that § is not susceptible to direct measurement in the sense in which 7? is. It ‘would be misleading, for example, to say that (nt) is the averages ured particle flux at the point r and the time #, for a mensuren average local flux implies simultaneous high-precision meusurenacnts of position and velocity (which is equivalent to momentum) and is (Tre-refore inconsistent with the uncertainty relation (3.1). Nevertheless sometimes helpful to think of S as a flux vector, especially w depends only slightly or not at all on r, 0 that an uecnrate determination can be made without impairing the usefulness of th concept of fix. Expectation Value. The existenco of the position probability «tensity P(r) takes it possible to caleulate what we shall call the crzre-etatio value of the position vector of a particle, which is defined as the veetat whose components are the weighted averages of the corresponiliny sx cat ponents of the position of the particle. ‘The expectat ne is the mathematical expectation (in the sense of probability theory) fer the result of a single measurement, or it is the average of the restaLtx of Jarge number of measurements on independent xyxtems of the tape die ‘cussed at the beginning of this section, We write the expectation value of ras + div SG) = 0 ) = [xP (ode = fUeOrwO,dr 5) ‘which is equivalent to the three equations (@) = Wate, ty) = Stade, (2) = eps80. 71 THE SCHRODINGRR WAVE EQUATION 25 where y is normalized. ‘The expectation value is # function only of the time, since ¥ and P depend on ¢ and the space coordinates have been integrated out. ‘The expectation values of any other physically meaningful quantities can be found in a similar way if they are functions only of the particle coordinate r. ‘Thus the expectation value of the potential energy is W) = [VG MP Cdr = JHE) VE edr (7.6) A quantity such as momentum or energy must, however, be expressed in terms of rand before a calculation of this type can be made. We assume tat it is possible to use the differential-operator representations given in Eq. (6.18) for this purpose, and justify this assumption with the help of the corresponding probability functions in Seo. 10 (for the energy) and in Seo. 11 (for the momentum). The question immediately arises, however, a8 to how such differential operators are to be combined with the position probability density P. ‘This question may be answered by imposing on the expectation values the reasonable requirement that, =F +m) in analogy with the classical energy equation (6.15). In terms of dif- ferential operators, this may be written C0 CMM an tis apparent that Eq, (7.7) is consistent with the wave equation (6.10) only if the expectation value is defined in the genefal case with the operator acting on y, und multiplied on the loft by J. We therefore obtain, for example, = [rte @- [1-meadve ca) Like Ba, (7.5, the aceon of Tas (7.8) equivalent tothe then com ponent equations coy =i fr%ar, y= in [ oie on=-a foe Ehrenfest’s Theorem. It is reasonable to expect the motion of a wave packet to agree with the motion of the corresponding classical 'P. Khronfest, Zeit f. Phi, 45, 455 (1027),26 QUANTUM MECHANICS (onan. 11 particle whenever the potential energy changes by a negligible amount ‘over the dimensions of the packet. If we mean by the “position” and “momentum” vectors of the packet the weighted averages or expec- ‘tation values of these quantities, we can show that the classical and quantum motions always agree. A component of the “velocity” of the packet will be the time rate of change of the expectation value of that component of the position; since (r) depends only on the time, and the # in the integrand of Eq, (7.5) is a variable of integration, this is Bern Gf] vevar= [rothans [avi ‘This may be simplified by substituting for the time derivatives of the wave function and its complex conjugate from Eq. (6.16) and canceling the V terms. de- ~i[ fu (-Bove ve) ~ | (- Revs ve) av] = B | ew - venie ‘The second integral can be integrated by parts: J[ (vere = — f (grad) - grad (ey)ir + f, (ov grad Yar. dA where the integral of the normal component of 2¥ grad j over the infinite bounding surface A is zero because a wave packet ¥ vanishes at great distances, A second partial integration, in which the surface integral again vanishes, results in SOD abe = JOY ar Thus Se = 2 [ ory —vepie Since (2) is seen always to be a real number from the structure of Eq. (7.5), Eq, (7.9) shows quite incidentally that (p,) is veal; this can also be shown from the second of Iegs. (7.8) when y represents a wave packet, bby means of partial integration. p.) 7.9)Spe. 8] THE SCHRODINGRR WAVE QUATION 27 In similar fashion we can calculate the time rate of change of a com- ponent of the “momentum” of the particle as Giro = ad [ 02 = -a( [og des [ee “) . -[ed(-z wwe ve)at [(- Rows ve) te --fe ‘le wy - a foe a dee <- 7.10) again substituting from the wave equation and integrating twice by parts. ‘Equations (7.0) and (7.10), together with their other components, are analogous to the classical equations of motion, ae Hak Bo gay They provide an example of the corrompondence prineiple, since they show that ® wave pucket moves like a elusieal particle whenever the expectation value gives a good representation of Uhe elnssieal variable; thisis usually the macroscopie limit in which the finite size and the internal structure of the packet ean be ignored. 8. ENERGY EIGENFUNCTIONS ‘Tho SohrBdinger wave equation (6.16) admits of consideruble sim- plifieation when the potential energy V(r) docs not depend on the time. It is then possible to express its general solution as a sum of products of functions of r and ¢ separately. Separation of the Wave Equation. We consider a particular solution, cf Haq. (6.16) that-can bo written ax n product: ¥( 4) = u(2)/(05 « general solution ean then be written asa sum of such separated solutions. — Tf we substitute into Ng, (6,16) and divide through by the produet, wo obtain hay. ac i(- Vue Veer) en) Since the left side of Hq, (8.1) depends only on ¢ and the right side only oon f, both sides must be equal to the sume separation constant, which we call B. ‘Then the equation for fis readily integrated to give 2 MW = Ce ®28 QUANTUM MECHANICS (Cua. 11 where Cis an arbitrary constant, and the equation for u becomes, [- zvet vo| ue) = ue) 62) Since Eq. (8.2) is homogeneous in u, the constant C’ may be chosen to normalize u, ‘Then a particular solution of the wave equation is Ue) = ure 3) Significance of the Separation Constant B. ‘The time-derivative operator given in Eq. (6.13) as a representation of the total energy may bbe applied to the ¥ of Eq. (8.3) to give ov att = By (64) An equation of the type of Eq. (8.4) is called an eigenvalue equation; y is said to be an eigenfunction of the operator that appears on the left, and the multiplying constant # thet appears on the right is called the corresponding eigenvalue. An energy eigenfunction, like the ¥ in Eq, (8.3), is said to represent a stationary state of the partie, since |y|? is constant in time. Equation (8.2) is also an eigenvalue equation. Tt states that u (and hence also y) is an eigenfunction of the operator (~ (A*/2m)¥? + V(@)] with the same cigenvalue . Tt is, of course, to be expested that y is an eigenfunction of this operator if it is an eigenfunction of the time- derivative operator, since according to the wave equation (6.16), the two operators are equivalent not only for separated functions of the form of Eq, (8.3) but also for more general solutions. ‘We now anticipate the discussion of the physical significance of eigenfunctions and eigenvalues that will be presented in Chap. III, and assume that the energy eigenvalues # are the only possible results of precise measurements of the total energy of the particle. It is then of interest to inquire whether or not physically interesting solutions w(t) of Eq, (8.2) exist for all real values of Z. An answer cannot be obtained until a specification of “physical interest” is found in terms of the bound- ary conditions that are imposed on u(r). ‘This specificntion, and the general character of the energy eigenvalues that are associated with various types of potential energy function V(x), are considered in the remainder of this section. The terms characteristic function and characteriatic value are often used in place of eigenfunction and efgonoalue,Sto. 8] THE SCHRODINGER WAVE EQUATION 29 Boundary Conditions at Great Distances. We have thus far encoun- tored two classes of wave functions: wave packets that are well localized and for which the normalization integral {|y|*ér converges, and traveling harmonic waves like the function (6.10) that have a constant magnitude at great distances so that the normalization integral taken over an infinite volume diverges. ‘The first class may be interpreted as represent ing particles tht, if free, are initially well localized, or that are restrained toa particular region of space by external forees derived from the poten- tial energy V(@). ‘The second class represents particles that are neither localized nor restrained, but travel through the region under consideration from one distant part of space to another; stich wave functions will be useful in describing the scattering of particles by a field of force.’ In either case, the wave functions are bounded at great distances in all directions. Continuity Conditions. ‘The time-independent wave equation (8.2) is 1 second-order linear differential equation in r. ‘Thus 60 long as V(t) is finite, whether or not it is continuous, a knowledge of the wave function and its gradiont along a surface makes it possible to integrate the equa- tion to obtain the wave function at any point. Ttis reasonable, therefore, to require that the wave function and its gradient be continuous, finite, and single-valued ut every point in space, in order that a definite physical situation can be represented uniquely by a wave function. ‘These requirements also have the consequence that tho position probability density P(e) and tho probability current density $(f), defined in See. 7, are finite and continuous everywhere. Boundary Conditions for Infinite Potential Energy. If V() is infinite anywhere, the appropriate boundary condition cun be established by a limiting procoss that starts from a Anite V and the above continuity conditions. Suppose, for example, that there is an infinite discontinuity in V across a continuous surface, so that the potential energy is finite on one side of it and + on the other, and we wih to determine the boundary conditions on u(e) and grad u at this surfwe, ‘The essential fentures of the problem are rotained if we replie the continuous surfuee by the plane that is tangent to it at the point of interest, und the continuously changing potential energy on one side of the surface by a constant poten- tial, which can, without loss of generality, be chosen to be zero since any ‘constant change in V is equivalent to an equal change in Z. We choose 1 Anothor possible class contista of wave functions that become infinite wt largo distances; however, thoso are not of physical interest, since wo have no reason to be oemed with particles for which the pasition probability density becoiney inde finitely lange in remote regions of spac,30 QUANTUM ABCHANICS the origin of coordinates at the point of interest and the z axis perpendicu~ lar to the tangent plane. ‘The wave equation (8.2) then separates in the three space coordinates, and the dependence of w on y and z is not affected by the discontinuity in V at the plane z = 0. We wish therefore to solve the one-dimensional wave equation ~ ERs v@u= Bu @5) where V(z) =0 for z <0, V(z) = Ve for 2 >0, and we eventually pass to the limit Vi—> +e. If we assume that 0 < E < Vz, the general solutions of Eq. (8.8) are u(z) = A sin ox + B cos az, 2<0, a= 7 - 4 [2mm] ule) = Cr+ De, 2>0, 8 + [=] ‘The boundary condition that w be bounded at great distances requires that we set D = 0. Then the continuity of u at x = O gives the relation B =O, while the continuity of du/de gives the relation ad = —BC. Since # becomes infinite when Ve does, and the solution for x <0 must be finite, the second relation shows that C becomes zero as Ve =, thus also making B zero; A is not determined from these relations, but might be fixed by normalization. ‘Thus the boundary conditions at o, surface at which there is an infinite potential step are that the wave function is zero and the component, of the gradient of the wave function normal to the surface is not determined. ‘The assumption above that # < Vo is evidently not a restriction since V, eventually becomes infinite. For # <0, the sine and cosine in the solution for 2 <0 are replaced by hyperbolic sine and cosine (which is permissible since the solution need hold only near 2 = 0), with no change in the final result. It should be noted that both P and S, vanish as 2+ 0 from the negative side, so that they are continuous at x = 0 even though du/dr is not. A boundary surface of this type represents a perfectly rigid, impene- trable wall, since in the anslogous classical situation a particle of any finite energy would have its x component of momentum reversed instar- taneously on contact with the surface. Energy Eigenvalues in One Dimension, Energy eigenfunctions that represent particles that are restrained to a particular region of space by the potential energy (frst class) are always characterized by discreteSeo. 8) ‘THE SCHRODINGER WAVE EQUATION a1 eigenvalues, while eigenfunctions that do not vanish at great distances (ecoud class) possess a continuous range of eigenvalues. ‘This may beseen qualitatively by considering the nature of the solutions of the one- dimensional wave equation (8.5). ‘We shall suppose at fit that V(r) becomes equal to some constant ‘value, which may be taken to be zero, for sufficiently large positive and negative z, and that H <0. A classical particle with this total energy 2 cannot escape to infinity, and indeed can exist in th region only if B is greater than or equal to the smallest value Vain of V(x). ‘The permitted form of the wave function for |2| large enough that V = 0 is evidently #21, where 8 = +(—2mB/A*)). These two solutions for large positive and negative x can be extended in toward some intermediate point, say = 0, by making use of the wave equation and the continuity conditions. At this point, u can always be made continuous by choosing the arbitrary raultiplying constants for the positive and negative « solutions appropri- ately. ‘Then @ little reflection shows that for an arbitrary value of B, they will not meet with the same value of du/dz. ‘There may, however, be particular values of # for which both u and du/de are continuous at == 0. Theconditions for this ean be seen in the following way. In the rogions in which B < V(x), du/de* i positive, and sow is convex towards the x axis, ‘Thus the two solutions that are continued in from + « have oppusitesigns for their ratios of slope to value, tidy, Jong as they are in regions for which # is always less than V. This is strated in Fig. 6 for the potential shown in Pig. 54; hoth choices of the sign of w aro shown for 2 <0. ‘The points at: which J? = V(z) are called the turning points (7P) of the elussical motion, xinco they nro the limits of the motion of a classical particle of energy JH, nt whieh the particle turns around or reverses its motion, At these points du/ia? = 0, and 1 has zero curvature, Clearly what is needed to make the tivo solutions join smoothly is 1. region in which E > ¥(e), in order that d'u/d2? muy be negative and 1 de coneave toward the z axis this may permit, the ratios of slope to valu to become equal. Figure 5e shows the tivo solutions cartied in until they ‘meet, but for a somewhat small value of , so that when the 1’ re made the sume at 2 = 0, the slopes are not eal (sold curves), oF when the slopes are made the sune at z = 0, the ware not equal (dotted eurve on left and solid curve on right). Figure 64 shows a xomewhal larger (les negative) velue of B, and Fig 5e a still larger value, ‘The values of32 QUANTUM MECHANICS [Ceap. 11 ‘and Vuu aro indicated on the u axis, and the turning points are indicated on the 2 axis, in the last three cases, Discrete Energy Levels. We sce then that an eigenfunction that satisfies the boundary and continuity conditions and that represents & particle bound by the potential energy V(z), can exist for the particular a eta) le. i 10 o| 4 . } = rn oO ‘a ted lara ——_— =e[>* = Se De = @ Tin Tao 5. () Poteau nergy fneton and () wlio fr Ire (9 aad (9 show that tithor the wave function or fs nope in Sacatinaoun st = = for value of tha are Salle (ee ngative) and lergen eapoctiely, than he onary egnval shown in dy value of H ilustrated in Pig. 5d. Tn analogy with the classical situation, ‘necessary condition that such an eigenfunction exis is that Vou <0, in which case Z lies between Vais and 0; as in the classical ease, this con- dition is also sufficient in one dimension although it is not in three dimen- sions (see Prob. 10, Chap. TV, and Sees. 9 and 15), Tf the poiential energy well illustrated in Fig. 60 ia euficiently broad cr deep, there will exist another eigenfunction corresponding to a larger ‘energy eigenvalue that is still negative. Figures Ga, 6, and c bow a series of wave functions analogous to those shown in Figs, 5 d, and ¢ for sue- cassively increasing (successively less negative) values of 7; both signSe. 8] THR SCHRODINGER WAVE BQUATION 33 of ware shown fort <0. ‘Thus Figs, 6d and 6b show th mnotior 4 ; 8 for the two lowest energy eigenvalue or energy lel of « wetweho by the potintia! well VG). Tt is easy to see by an extension of the fore. going qualitative arguments that if there are any higher disrete energy levels, exch vigenfunetion has one more node than that i the next luvvest eigenvalue, : eee ‘Thus for a potential energy that approaches a finite constant y 10 inite constant value as, #— £, thore may be a finite number of diserete energy levels, ot in Fra, 6, Solutions for rufciutly rod or dsep potential and larger (lem negstive) values ‘off thats thom abo tw Hig. 6.4? ineroaaer a gun from (a) to @) to.) and i om, tlgonvaliie ia 0), ava Uo wave funtion end ia slope can both be continuove at = 0. some cases an infinite number Gf V(2) falls off slowly enough for large 21), depending on V(r) and the mass of the particle. However, V(@) — +2 ax x—> +, an argument like that given above shows that thore will always be sn infinite number of discrete energy levels; apart from atl Continuous Energy Eigenvalues. Its possible to find eigenfunctions tha obey the boundary and continuity conditions for all-energy eigen- values that exceed the smaller of the two numbers V(-+ ) and V(~ ©). 1, for example, the potential energy hus the form illustrated in Fig 6a, then solutions of the wave equation ean be found for all positive values of B,Thixix bocwuse the solutions for large [| are of the form Asin aj] + B cos az, a= + (ey 8.8)34 QUANTUM MECHANICS [Omar 1 and there is no reason why both terms should not be kept. ‘Thus it is always possible to adjust the phase of each of the wave functions for large [sl (which is equivalent to adjusting the ratios A/B for the solutions for large positive and negative 2) so that they join together smoothly when continued in to z = 0. ‘The classical terms periodic (or multiply periodic) and aperiodic are sometimes used to designate the particle motions associated with discrete and continuous energy eigenvalues, respectively. Discrete and Continuous Eigenvalues in Three Dimensions. We shall assume without further discussion that all the foregoing results can wo (Pim) to ve) 4 % oe fa) 6) Fro. 7, Onedimensional xquare wel potential wih (0) pafedy ied walle and) faite potential we, be taken over in a natural way for the three-dimensional wave equation (82). We can expoct that if Vie) +© as r—+ © in all directions, there will be an infinite set of discrete energy levels extending to += If V(0) is bounded as r -> © in some direction, there may be a finite or an infinite number of discrete levels, depending on the form af V. In this case, the discrete energy levels cannot exceed the smallest value that Vs) bas in any direction. For values of # larger than this smallest V(«), the energy eigenvalues cover a continuous range extending to +, 9. ONE-DIMENSIONAL SQUARE WELL POTENTIAL As a simple explicit example of the calculation of diserote energy levels of a particle in quantum mechanies, we consider the one-dimensional motion of e particle that is restrained by reflecting walls that torminate a region of constant potential energy. Two simple types of potential energy are considered. Figure Za shows a situation in which V(z) = 0 for a a, corresponding to per- fectly rigid, impenetrable walls at the points z= ta. In Fig. 7) the increase in potential energy at the walls is abrupt, but finite, so that V(@) = Vo for |2| > a; because of its appearance, this is often called a ‘aquare well potential. ‘The motion of a classical particle with total energyS10. 9] THR SCHRODINGER WAVE EQUATION 35 less than Vo is the same for both these potentials; but as we shall see, ‘the quantum-mechanical behavior is different. In general, an abrupt finite increase in potential energy at the boundaries of a region forees particle toward the interior of the region. Such a potential may be thought of as a limiting ease of a potential of tho type shown in Fig. 5a, for which the fore —dV /dz is always direeted in toward x = 0. ‘The force associated with a square well potential is zero except at the boundaries, so that the particle is acted on by no force except a sudden impulse directed toward the origin as it passes the points x = ka. Perfectly Rigid Walls. Tt was shown in Sec. 8 that the wave function ‘must vanish at the points 2 = a, when tho potential energy has the form shown in Fig. 7a, From Eq. (8.6) the wave equation for |z| a ‘The wave equation in this region is ae dtu — EP vue Bw which has the general solution for F < Vo (bound states) oa) we acem spe, aa s/f Bl ea) ‘The boundary conditions at z = +c discussed in Sec. 8 require that we set D = 0 if Eq. (9.3) is to represent the solution for ¢ > a, and C = 0 if the solution is for 2 < —a. ‘We now impose on the solutions (9.2) and (9,8) the requirements that wand du/dz be continuous at z= a. Asin aa + Boosaa= Crt, aA cos aa ~ aB sin aa = —B0e* ~Asinaa + Beosaa=De*, aA cos aa+aB sin aa = BDe-® from which we obtain 24 sinaa = (C~D)e*, 2a cos aa -A(C — D)e™ (9.4) 2B cos aa = (C+ Die, — 2aB sin a AC + Dye (9.5) Unless A = 0 and C = D, Eqs. (9.4) have as their consequence acot aa = —8 (9.6) Similarly, unlew B = O and C = —D, Bas. (9.5) give atan aa = 8 9.7)Bro. 9) THE SCHRODINGRR WAVE EQUATION 37 Now it is impossible for Eqs. (9.6) and (9.7) to be valid at once, sinee ‘on elimination of f this would require that tan? aa = —1, which in tum ‘would make « imaginary and 6 negative, contrary to Bq. (8.3). Also, we donot want A, B,C, and D all to vanish. Thus the solutions may again be divided into two classes: For the first class A=0, C=D, ond atonan=6 ‘and for the second class B=0, C==-D, and acotaa=—6 Energy Levels. ‘The energy levels are found from numerical or graphical solution of Eqs. (9.6) and (9.7) with the definitions for « and 6 i pibtangs 1 t » { t { L. 4 i] % z 2 3 4 5 2 Fro, 8. Graphien! solution of Ra {he Gt to ayant of y= E 1) for three valuoy of Vent; vertical dashed lino aro given in Eqs. (9.2) und (9.3). A simple graphical method for effecting this solution is deseribed hero, since it shows quite clearly the way in which the number of diserete levels depends on Ve and a. We put = aa, » = Ba, Whence Iq, (9.7) bocomos € tan & = 9, with, 2mVea? ete ee Since £ and 9 are restricted to positive values, the energy levels may be found in this case from the intersections in the first quadrant of the curve of &tan £ plotted against &, with the circle of known radius (mV a?/24)8, ‘Tho construction is drawn in Fig. 8 for throe values of Vyat; for each of the two smaller of these values, there is one solution of Eq. (9.7) and for the largest there are two.38 QUANTUM MECHANICS [Cua 1 ‘Figure 0 is a similar construction for the solution of Haq. (9.6) in which the energy levels are obtained from the intersections of the same eireles with the curve of —£ cot fin the first quadrant. The smallest value of ‘Vea? gives no solution, and the two larger values each give one. ‘Thus the three increasing values of Vsa* give altogether one, two, and three energy levels, respectively. Itis clear from Figs, 8 and 9 that for a given particle mass, the energy levels depend on the parameters of the potential energy through the eom- bination Vea’. For Voa* between zero and x%*/8m, there is just one coh vay roo é ia, 8.” Graphio solution of Eq (0.0) fo trea values of Vu ‘int aiymptote of y =e oot & ial dashed line i tho energy level of the first cluss; for Vea" between +°2%/8m and four times this value, there is one energy level of each class, or two altogether. As Vea? increases, energy levels appear successively, first of one elass and then of the other. Tt is not dificult to see from Eq. (0.2) that when ordered according to increasing eigenvalues, the nth eigenfunction has n — 1 nodes. Parity. It follows from the foregoing discussion that the eigenfune- tions of the first class are even with respect to change in sign of z (u(—2) = u(z)], while the eigenfunctions of the second class are odd (u(—2) = —u(2)]. This division of the eigenfunctions into even and odd types is not accidental, and will now be shown to be a direct consequence of the fact that the potential energy function V(z) is symmetric about x = 0. If we change the sign of z in the wave equation (8.8) - EMO + vena = Bue (9.8)Sao, 8] THE SCHRODINGBR WAVE BQUATION 39 and if V(—a) = V(@), we obtain — BG + ves) = Bu(-2) ‘Then w(2) and u(—z) are solutions of the same wave equation with the same eigenvalue Z. Unless there are two or more linearly inde- pendent eigenfunctions that correspond to this energy evel, these two solutions ean differ only by a multiplicative constan cule). 9.9) Changing the sign of = in Eq. (0.0) gives u(z) = eu(—z). From these ‘ovo equations it follows at once that u(-a) or #1 ‘Thus all such eigenfunctions of a xymmetrie potential are either even or odd with respect to changes of sign of z. Such wave functions are said to have even or odd parity. If an eigenvalue has more than one linearly independent eigenfune- tion, these eigenfunetionx need not have a definite parity: they need not be even or odd. However, wo ean easily see that linear combinations of such eigenfunctions ean be found such that each has even or odd parity. Suppose that an cigenfunetion u(r) does not have a definite parity. Tt ceun always be weibten u(e) wz) + rz), where us(2) = Hu(e) + u(—z)] ix even and u,(2) = ¥[u(e) — u(—2)] odd. Then if the wave equation (0.8) is symmetrie, we ean write it, as 2 due 1 dus 2 — GE + VY - 1 - EGS + V0 (9.10) On changing the sign of x in Hig. (9.10), we obtain — EGE E Bet fs — ~ u=0 (0.11) Addition and subtraction of Hqs. (9.10) and (9.11) shows thut , and uy are separately solttions of the wave equation with the sane eigenvalue Z. A Simplified Solution, Knowledge that the solutions possess a defi nite parity sometimes simplifies the determination of the energy levels, singe then we need only find the solution for positive x. Tiven solutions have zero slope und odd solutions have zero value ut x = 0. Tf, for40 QuaNrUM MECHANICS (Crap. example, we wish to find the even solutions, Eqs. (9.2) and (9.8) can be replaced at once by ‘u(2) = B cos az, Oa Instead of making both u and du/de continuous at x = a, it is enough to make the ratio (1/u)(du/dz) continuous at z= a, since the normalizing constants B and C are eliminated thereby. ‘This gives Hq, (9.7) at once. Similarly, the odd solutions are u(z) = Asin ax, Oa ‘Then continuity of (1/u)(du/dz) at x = a immediately gives Eq. (9.6). Problems 1, Use the arguments of Seo. 6 to set up « differntial equation for y that involves ‘aovond time derivative of y, in the caso of a free particle, Disquss any solutions that ‘this equation has that are not shared by the free-particle Schradinger equation. ‘2 Show directly that (p.) ia real for a wave packet. 8. Show that for a throo-dimensional wave packet $e - 2) +p. 4 Caloulate tho energy levels and plot the eigenfunctions for the three bound states in the potantial of Fig, 7) when Vat = GAY/m, Compare with the first three ‘states for the potential of Fig. 7a. 5, Disoass the relation between the energy levels for the potential of Fig. 7b and ‘thoae for the potential: Vie) = +0, 2 <0; Via) = 0, 0<2 a 6. Show that if tho potential onergy V(r) is changed by a constant amount every~ where, the wave functions are unchanged. What is the effect on the energy cigenvalues?CHAPTER IL EIGENFUNCTIONS AND EIGENVALUES In Chap. TI, the Schrodinger wave equation was developed and applied to usimple problem, Some discussion of the physical interpreta- tion of the wave function was given in Sec. 7, ‘This relied for the most: part on the computation of average or expectation values of operators that represent various physical quantities. In addition, however, a position probability density was introduced. ‘Thus at the present point We are in a position to calculate from the wave function any property thut depends on the spatial distribution of a particle (for example, the ‘mean-square deviation of its position from the average), whereas we can only caloulate average values of operators other than position. Clearly ‘what is necced to round out our interpretation of the wave function is a ‘means of computing probability funetions for operators in general; this would incidentally enable us to obtain the expressions for expectation values used in See. 7 without separate assumptions, In this chapter, we fist set up three physical postulates from which complete interprotation of the wave function can be derived, and then apply them to a discwssion of the totnl energy and momentum of a particle, and to an illustrative problem. ‘The interpretation presented in Sec. 7 ean then be seen to be a special ease of the more general treatment given here, 10, INTERPRETATIVE POSTULATES AND ENERGY EIGENFUNCTIONS We start with the wave function ¥(), which ix a solution of Eq. (6.16) and describes the motion of a particle thut has mass m and poten- tial energy V(t). From this wave function we wish to obtain as complete a description of the properties of the motion of the particle as is eonsistent, with the uneertuinty relations disused in See. 3. ‘Dynamical Variables as Operators. We first postulute that each dynamical variable that relates (o the motion of the particle ean be represented by an operator. ‘The operator may bo vimply a multiplication operator stich as r for the position, or it may be « differential operator such us 1h grad for the momentum. With each operator ean be assoviated an oigenvaluie equation, defined a2 QUANTUM MECHANICS (Cuan. mt near the beginning of See. 8. Thus with the operator % may be associ- ated the equation Buy = chy (20. where ti is the eigenfunction of 9 corresponding to the eigenvalue w. ‘Our second postulate is that one or another of the eigenvalues w is the only possible reeult of a precise measurement of the dynamical variable represented by 0. This implies that the eigenvalues of all operators that represent physically significant variables are real numbers, ‘Expansion in Figenfunctions. We assume that all the eigenfunctions of any dynamical variable constitute a complete set of functions in the sense thet an arbitrary continuous function ean be expanded in terms of them. This is a mathematical, not a physical, assumption, and will be discussed further below in connection with energy and momentum eigenfunctions, Suppose now that a particular wave function y is expanded in terms of the eigenfunctions uy of the operator . We adopt the statistical inter~ pretation of y given at the beginning of Sec. 7, sccording to which there area large number of identical, nonoverlapping regions of space, in cach of which is » partiole described by y. We then make measurements of the dynamical variable represented by @ on each of these particles. ‘Our third physical postulate is that the number of meamurements that result in the eigenvalue w is proportional to the square of the magnitude of tht coeficient of tu, in the expansion of y. This enables us to associate probability function with any dynamical variable. A corollary is that wwe are certain to measure a particular eigenvalue w only when the wave function thet describes the particle is the corresponding eigenfunction tur Rather than develop the consequences of these postulates for an arbitrary operator 8, we consider here tho total energy of the particle, and in Sec. 11 the momentum. Most of the results obtained are readily applicable to other physical operators. The Total Energy Operator. According to the uncertainty relation (8.3), @ precise measurement of the total cnergy of a particle cannot be made in a finite length of time. ‘Thus if the total energy is to have @ definite value, it is essentiel that the potential energy V(r) be inde- pendent of the time, Then the operator —(ht/2m)V? + V(), which is shown in Sec. 8 to be equivalent to the total energy operator 14 3/4t, has cigenfunetions u(r) that need not involve the time. ‘The energy-cigen- value equation is Eq. (8.2), [- Bet veo] esto = ut 10.2)nc. 10] BIGENFUNCTIONS AND EIGENVALUES 43 where the eigenfunction wx(t) corresponds to the eigenvalue E, and obeys the boundary and continuity conditions of Sec. 8. As discussed in Sec. 8, the energy eigenfunctions ean be divided into two classes: those which are well localized and are associated with discrete igenvalues, and those which remain finite at great distances and possess continuous range of eigenvalues. ‘Normalization in a Box. It is often desirable to be able to treat these two classes on the same basis; this ean be accomplished by enclosing the particle under investigation in a box of arbitrarily large but finite volume. ‘The simplest physical situation to which this approach is applicable is ‘one in which the walls of the box are perfectly rigid, so that, as shown in Sec. 8, the wave function vanishes there, Then the discussion of Sec. 8 shows that all the cigenvalues are discrete. If the box is large in com- parison with the dimensions of physical interest in the problem, the eigenvalues that were discrete in the absence of the box are practically unaffected, since before the walls were introduced the wave functions were extremely ‘small there. Also, the eigenvalues that were continuously distributed in the absence of the box are very closely paced; this is shown explicitly for a free particle in Sec. 11. It is more convenient to assume that the wave functions obey periodic boundary conditions at, the walls of the box than that they vanish there, since it is then possible to get a simpler description of the momentum ‘eigenfunctions (see Sec. 11). We choose the finite region to be a cube of edge length L centered at the otigin, and require each wave funetion to have the same value at comesponding points of opposite faces of the cube, and the samo derivative normal to the wall. ‘These boundary conditions make the otherwise continuous eigenvalues discrete, since ‘the phase of the eigenfunction at great distances is no longer arbitrary {seo the discussion of Hq. (8.6)]. Ax with the rigid-walled box, the pres- ence of the walls has a negligible effoct apart. from imparting disereteness to the otherwise continuons cigenvalnes and providing x finite volume in which these wave functions ean be normalized; we hall continue to te the word “contimuous” to describe these functions, even when box normalization ix used Orthonormality of Energy Eigenfunctions. ‘Ihe intogral flus(c)/lr, which converges in any ense for the discrote sot of eigenfuneti ‘verges for all eigenfunctions when they are normal finite volume 14, ‘The coefficient of ue can then he chosen x0 that: this integral is equal to unity, and the wx(r) are normalized. ‘We now show that the eigenfunctions corresponding to two different, eigenvalues and i’ arc orthogonal; that is, that the integral of the product of one of them and the complex conjugate of the other over the“4 QUANTUM MecHANICS (nae, common domain of the functions is zero. From Eq, (10.2) we have that Gr (2) satisfies the equation [- Boe ve] tele) = F'ar(e) (103) ‘where wo assume in accordance with our physical interpretation that J” is real; this is verified below. We multiply Hq, (10.2) by dw and Eq. (20.3) by us, integrate over the volume L*, and take the difference between the two resulting equations. The V terms cancel, and leave -£ i, (Ge ¥!up — usV%ay)dr = (B — BY) f ayusir (104) ‘The integral on the left side of Eq. (10.4) can be transformed by Green's theorem into a surface integral over the surface A of the cube, Jae V%e = wsV'de)dr & f div (ay grad us ~ ue grad ay)dr = I (dy grad us — us grad dy) yA (10.5) where the subscript m designates the component of the vector in the Girection of the outward normal to the element of area dA. Since the ‘inmposition of periodic boundary conditions gives each wave function and its normal derivative the same values at corresponding points of opposite faces of the cube, the outward normal derivative hus opposite signs om opposite faces, and the surface integral in Eq. (10.5) vanishes. ‘Then Eq. (10.4) tells us that if B x H’, wy and wy are orthogonal. An energy eigenvalue Z is said to be degenerate when two or more lin- early independent eigenfunctions ws, ws, . . . correspond toit. Orthog- onal linear combinations of degenerate eigenfunctions cam be found in many different ways. For example, ua = aru: + art: can be mado orthogonal to ws by choosing the constant coefficients a; and az such that 1 _ Saude a [ua]*dr this choice does not interfere with the normalization of we and 1s is still an energy eigenfunction with the eigenvalue #. Evidently the choice 1 1¢ is apparent that this proof of orthogonality can be applied to a diaerde at of ‘eigenfunctions even though the box is not introdused, since the w'a vanish rapidly at sweat distances and the surface integral, which ia thon over a aphero of infinite radius, is ero. The continuous wet of enorgy cigenfunctions ean also be treated without ‘using the box normalization (as is the continuous set of momentum eigenfunctions in Sec. 11). S00 B. C. Kamblo, ‘The Fundamental Principles of Quantum Mechnies”” (MoGraw-Hill, New Yori, 1987); Bee. 90 of Komble's book prosenta a discussion of this problem and references to the original papers.Sue, 10) BIGENFUNCTIONS AND EIGENVALUES 45 of orthogonal linear combinations is not unique. By an extension of this procedure, all the energy eigenfunctions ean be made orthogonal to each other even though some of the eigenvalues are degenerate. Such a sot of eigenfunctions, each of which is normalized and orthog- onal to each of the others, is called an orthonormal set of functions. We specify an orthonormal set of nondegenerate energy eigenfunctions by the relation Jae(e)uale)dr = baw (10.6) where bra is the Kronecker 8 symbol that equals unity if # = E’ and is zero otherwise. If there is degeneracy, Eq, (10.6) must be replaced by Jtro(eus(t)dr = Brxbur 0.7) where the index s distinguishes between orthogonal degenerate eigen- functions. Tt is often convenient to omit explicit mention of # and use Rg, (10.6) for degenerate situations as well, in which case the index s is implied, Reality of the Energy Eigenvalues. We can now sce dirvctly that I is areal number, as hus beon assumed, We multiply iq. (10.2) by (2) and integrate over the box of volume J, Tf ux is normalized, the result ix coe & f seve + J V()junltde which may be expressed in terms of expectation values (1/2m)(p") + ui) 4 Aad = > Axd)Eus(e) (10.16) ‘Because of the oe of the us, Eq, (10.16) is equivalent to ind Ast) = BAW) which may be integrated at once to give sty Ast) = Aalto)e™ Note that P(Z) = |Ax(t)|* = |Aa(to)|*is constant in time. ‘Thus if ¥(ry) is known at the time ¢ = fo, the solution at any time ¢ is given in terms of Eqs. (10.15) and (10.17): ve) = F Aatvoe™ eal) (10.18) Ankle) = J inlet gaddr! (10.17) or eu=19 Ve = [ [YS aeteruate™ P| vlerwadier (10.19) ‘The solution (10.18) is a linear combination of the separated solutions (8.8) obtained earlier. 41, MOMENTUM EIGENFUNCTIONS ‘The eigenfunctions of the linear momentum operator —iA grad pro- vide second instructive example of the application of the general ideas developed at the beginning of the last section. ‘They also are of con- siderable usefulness in solving problems of physical interest. Form of the Eigenfunctions. The momentum eigenfunctions are lutions of the three eigenvalue equations ih grad tig(e) = pup(t) aySn, 11]—-RIGENFUNCTIONS AND RIGENVALUUS ry or: —rdue) = pay), -m®2 a6) = pauple) Ju = parte, Fx) = pat a 0 Z uy(t) = pale) "They have the form t4(0) = C exp HP where C is a normalization constant. It is convenient, as in Sec. 6, to change from the momentum vec- tor p to the propagation vector k = p/h, and rewrite the momentum eigenfunctions: welt) = C exp dk-r (11.2) ‘Those are eigenfunctions of the momentum operator with the eigenvalues ne Box Normalizatioa. As with the energy eigenfunctions dinouned in See. 10, we ean restrict the domain of the (2) to un arbitrarily large but finio eubieal box of volume 1 entered at the origin, at tho walls uf which the functions obey periodic boundary conditions, Then te is normalized if L-t, Also, kt is no longer un arbitrary real vecto ita components ar rercted to the vation bya, = (13) where na, ny, and ng ate positive or negutive integers oF zero. ‘The spa ing of neighhoring i veetors ancl of their energy eigenvalies W'k?/2m can ‘be made ax small sx desivvd by making J sulficiently largo. It is tw note that the momentum cigenfunctions (11.2) ‘cannot exist within a box that lau perfeetly rigid walls, wines these cige Funetions do not vanish anywhere. "This ix analogous to the classical situation in which the momentum of a particle that is reflected from a rigid wall in not conserved. On the other hand, the eubieal box with periodic houndury conditions ix ocquivalont to a situation in whieh the entire infinite spnce is divided up into adjacent eathess and all wave fune= tions are periodic throughout xpace with the poriad J, along each of the three cartesian axes. If tho periodicity of the spaco ix carried over to the analogous classical situution, a particle passing through a wall would be equivalent to one that strikes that wall and appeans at the correspond ing point of the opposite wall with its momentum veetor unchunged. ‘The orthonormality of the momentum eigenfunctions ux(e) = Et oxp ler aa)50 QUANTUM MECHANICS (Cusp. I is readily established. For integration over the volume L* [amine fueron [Sar | = Bat hstyDhaly Bic (1.5) where the 4's are Kronecker 8 symbols and uso is made of Eqs, (11.8). Orthogonality could also have been shown by the more general method ‘used in See. 10 for the energy eigenfunctions [see Eq. (10.4)]. ‘The Dirac § Function, It was stated in See. 10 that continuous sets of eigenfunctions ean be handled without introducing the box with periodic boundary conditions (which has the effect of making the set discrete with ‘an arbitrarily small spacing of eigenvalues). This can be shown explicitly for the momentum eigenfunctions with the help of the Dirae 6 function,’ which ean be defined by the relations ae) = 0 ifr x80, faez)de= (1.6) ‘where the region of integration ineludes the point x = 0, An equivalent definition is that for an arbitrary function /(e) that is continuous at 2 = 0, the equation Sf@)6@)dz = $(0) aL7) is valid, where again the integration includes the point » = 0. Tt is apparent from & comparison of Eqs. (11.6) and (10.11), or of Bas. (11.7) and (10.10), that the bracketed quantity in Hq. (10.10) can ‘be expressed in terms of & functions: Jaren = We — 2)ily — vile — ¢) =e le ~ W') (11.8) Comparison of Eqs. (11.8) and (10.6) shows that the closure property is a kind of orthonormality of the eigenfunctions with respect: to summa- tion over the eigenvalues. A Representation of the 5 Function. ‘The definition (11.6) or (11.7) shows that 6(z) is an exceedingly singular function, Tt may be thought of qualitatively as being zero everywhere except at x = 0, and being 80 large there that the area between it and the © axis is finite and equal to unity. More formally, it can be represented in a number of different ‘ways as the limit of « sequence of analytic functions. ‘A. particular representation that is quite useful involves (sin g2)/2 1s. function of 2, where g is & positive real number. This has the value a/x at x = 0, oscillates with decreasing amplitude and with period 2x/g as [2| increases, and has unit integral from z= —« tox = +e indepen- +P. A. M. Dirae, “The Principles of Quantum Mechanics,” 2d ed, Seas. 20, 21 (Oxford, New York, 1985).pe, 11] EIGENFUNCTIONS AND EIGENVALUES 51 dently of the value of g. ‘Thus the limit of this function as g —> © has all the properties of the 5 function: it becomes infinitely large at z = 0, it has unit integral, and the infinitely rapid oscillations as |2| increases mean that the entire contribution to an integral containing this function comes from the neighborhood of = 0. We can therefore put sin ge (11.9) Normalization in Terms of the § Function, ‘The representation (11.9) of the 6 function can be used to set up an orthonormality integral like that given in Eq, (11.5), where now we do not impose the box normalization ‘but allow the momentum eigenfunctions to have the form (11.2) over all space with all real vectors k. ‘The integral [aj(t)ux(t)dr is the product of three integrals, each of which can be expressed in terms of a 8 function: [dont ae = tim [?, dmtor de 52) = lim = lim = Brbl(ke — 1) (11.10) ‘Thus the momentum cigenfimetions in infinite space can be written ue(e) = (Br8)-} exp dk qa) in which case the orthonormality relation becomes Si r)me(ltr = (ke — 12)8(ky — Ly)8(he — Le) = (Ke — 1) (11,12) See, 12 that the box andl 3-funetion normalizations Tt will be shown, of the momontum eigenfunctions give the same final result in a typical roblem. PMSome Properties of the § Function. Tt is important to note that, Docatiso of is singular character, the 6 funetion cannot he the ond result of acaleulation, ancl hax meaning only so long a 2 subsquient integration vor ils argument is cartied out. With this understanding we can write down somo relations betwen & functions. Hr) = (=a) 3) = -2(-2) 8) = 0 8) = —a(2) (L8) Slax) = a-'a(2), a>0 (et — a2) = 2ar'fale — a) + H(e + 0)], a>0 J8(a ~ 2)8(2 ~ b)de = (a — b) A@)a( — a) = f(a)a(z — a) Here, a prime denotes differentiation with respect to the argument. +B. ALM, Dias, op. typ 7.52 QUANTUM MECHANICS ema, 1 ‘Bach of the first six of these equations can be established by multiply- ing both sides by a continuous, differentiable function f(z) and integrating over z. For example, the fourth of His. (11.13) gives i fl@)2(@)de = — i ae) 2 testes JONI) + af wld = — fHl2)e@)de where the boundary terms that result from the partial integration vanish. ‘Thus 28'(c) has the same effect when it is a factor in an integrand as hos —&(z). Similarly the seventh of Eqs. (11.13) means that the two sides give the same result when multiplied by (a) or f(b) and integrated overa orb. The last equation is verified by integrating both sides over either tora. Closure, ‘The closure property of the momentum eigenfunctions, ‘with both box and é function normalization, can be established without ‘the help of the completeness assumption that was made in Sec. 10 for the energy eigenfunctions. With box normalization, the expression analogous to the left side of Eq. (11.8) is Jarmo =r F FF creirntmiranenenore ‘This is readily evaluated in the limit of large Z, in which ease the sum- mand changes by a negligible amount as each n changes by one unit, Wo can then regard n, as a continuous variable and replace) by [ledne= C/2a) [di We thus obtain Jaen) oo eI [7 [TL [1 eee rn icoide did, = ie — 28 — ¥)8— #) = He Py (ALI) oon making use of Eq, (11.10). A similar calculation ean be carried through, using the 6-function normalization, in which case we obtain from Eqs, (11.11) and (11.10) Ste(e ue (tare = Sf Jae(e)ue(t) dhe dky dks = a(r — x’) (11.18) __ The closure relation (11.14) or (11.15) shows thet the momentum eigenfunctions are orthonormal with respect to summation or integration over the eigenvalue k as well as with respect to integration over the posi- tion vector 1.See. 11] BIOBNPUNCTIONS AND BIGENVALUBS 53 Expansion in Momentum Higenfunctions. An arbitrary continuous function Y(t) can be written in terms of the 6 function as, VC) = Joleyate — ede (a1) If we substitute the left side of Eq, (11,14) in place of a(r — r’) in Eq. (11.16), we obtain ve) [ H0Y me duatedar’ =P Are(e) An = SiC’ W(e)dr” Similarly, if we substitute for a(r — r') from Eq. (11.15), we obtain Cr) = Jl") ticle ua e)dredr? = fAntun(e)dre (11.18) with the same expression for Ax, Equations (11.17) and (11.18) show that it is possible to expand an arbitrary function in momentum eigen- functions that are normalized either in a box or by means of 6 functions." Probability Function and Expectation Value. The momentum probability function associated with a normalized wave function ¥(«) is proportional to |As|?. The proportionality factor is unity, since if we put @1.17) P(e) = |Aul? (9) ‘it is easily shown in analogy with the summation of Tq. (10.12) that Prd —1 ond [Pedra = 1 (11.20) for the box and é-netion normalization, respectively. The expectiation value of the momentum when box normalization is need is @) = AY EPG) = AY f ktxly(eer f mele Hed? (11.21) From tho complex conjugate of Bq. (11.1), we ean replace keix() by 4 grad tig(e). ‘Then the first intogral in iq. (11.21) ean be integrated by parts and the surfue integral vanishes beeauso of the periodio boundary conditions on y and wy. Thus with the help of Fa. (11.14), Iq. (11.21) becomes 0) = mY f o400 grad yloer f mcenveenae = ih {00 arad VeoNs(e = eye a = =it{¥@) grad yin)tr : “thet rela, wih not rigorously entabnd her, aro equivalent tothe math rateal theorent on tho expasabliy of funelons In Fourier wee ad Rouser tapics QUANTUM MECHANICS omar. UI ‘This is in agreement with the second of Eqs. (7.8). ‘When é-function normalization is used, the details of the calculation are very similar to those given above, except that the surface integral that results from the partial integration is over a sphere of infinite radius; it is zero because p becomes vanishingly small at great distances. Thisis consistent with the supposition that y is normalized; otherwise neither SP (dra not (p) have any physical meaning. The result of the calcula ‘ion in this case is the same as qs. (11.22) and (7.8). 12, MOTION OF A FREE WAVE PACKET IN ONE DIMENSION The motion of a free particle (no external forves) in one dimension is described by the Schrédinger wave equation (6.8). The siudy of this ‘motion provides an interesting application of the expansion techniques developed in Secs, 10 and 11. As a first step, we find the minimum value of the uncertainty product given in Eq. (3.1), and the possible forms of the one-dimensional wave packet that correspond to it, all at a definite instant of time. ‘The structure of this minimum packet is the same whether or not the particle is free, since this form can be regarded simply 5 an initial condition on the solution of the Schrodinger equation for any V. However, the analytical work involved in finding y at other times is especially simple in the forco-free ease. ‘The Minimum Uncertainty Product. In order to find the minimum ‘value for the uncertainty produce Az- Ap, we must first define what we mean by Az and Ap. While many expressions are possible, the simplest to handle analytically is the root-mean-square deviation from the mean, ‘where the word “mean” implies the expectation value of Sec. 7. (aa)? = (e — (2) = (2%) — (2a(e)) + (2) = — GY ya ty (ap)? = ((p — (P))") *) — py : Here the equalities follow directly from the general definition af expecta tion value given in Seo. 7. If now we put ex2-@), per—@=-a[£-C] can (a2)*(ap) = [" darvde f” gatvds = [enends ("None — 02.8) ‘The transformation of the a integral in Eq. (12.8) is obvious; the similar transformation of the 8 integral follows from a partial integration when A.W, Heisenberg, “The Physical Principles of the Quentum Theory,” pp. 17-19 (Wniversity of Chicago Press, Chicago, 1930), thenSeo, 12] RIGBNFUNCTIONS AND BIGENVALUBS 55 we remember that y it a normalized wave packet, which vanishes at z= te, ‘The inequality dz BO Sipe Jlo- oft where all integrals are from z = 2 to +, equality is applicable only if f = yg, where + inequality we obtain at, oneo SdseSlalteee & [SFod|* If now we replace / by ay und g by By, Hq. (12.3) becomes (ax)*(ap)? & | (Gd) (BH)dex|* = | Pasiyda|* (2.4) ‘The last term in Hq. (12.4) ean be written [JPR (o8 — Be) + H(aB + Bex) Medel? = HJPlap — Badite|* + HfP(ap + Bayete|® (12.5) ‘The cross term in the produet that is omitted on the right side of Eq. (12.5) ean bo seen to vanish when use is made of the relation Tay Ce = JWaBY dv = [BY op) de = Sbsiav de, which is obtained by using pa is real, Now from Wig. (12.2) obviously true, and the constant, From this eygration and remembering that a (8 = feay = fife — cov] = ina (120) We thus obtain from Bas, (12.1), (12.6), and (12.6) (antiapyy Bor areay BAA (27) where the equality ean hold only if the second erm on the right side of Bq. (12.5) is mero, ‘This is the precise expression of the Heisenberg iy relation (3.1), when the uncertainties Ar and Ap are defined asin By. (12.1) Form of the Minimum Packet. 11 follows from the foregoing deriva tion that the minimum uncertainty product i ablained only when two conditions sre fulfilled, oy = (12.8) SMab + Baye = 0 (12.9)56 QUANTUM MECHANICS [Cnae, I ‘Equations (12.8) and (12.2) give us a differential equation for y &-[5e-on+@]y which is readily integrated to give G2) = Nexp [es (-G@y+ se] (1210) where N is an erbitrary constant. ‘Equation (12.9), with the help of Eq. (12.8), becomes, (t+) [mv which evidently requires that be pure imaginary. ‘Then since we want, Eq, (12.10) to represent a wave packet for which the integral of |¥/ con- ‘verges, 7 must be negative imaginary. The magnitude of the constant V ‘can now be fixed by normalizing y. Siwltdz = 1 Similarly, 7 can be determined by requiring that Se ~ (2))*\p ltd = (42)* ‘The integrals are readily evaluated, and lead to the normalized minimum wave packet Y(e) = [2e(A2)'I-t exp [- ae +i (2) Momentum Expansion Coefficients. ‘The one-limensional momen- ‘tum eigenfunctions analogous to Eqs. (11.4) and (11.11) are 0 tu(z) = Lote (12.12) for normalisation in a one-dimensional “box” of length L, and sua(z) = (2m) tem (12.18) for éfunction normalization. Since for a free particle the wave equation has the simple form of Eq. (6.8) ee ey ne - EY (02.14) the momentum eigenfunctions are also eigenfunctions of the energy.' The converse is not necessarily true, since there ar tro solutions of Faq. (12.18) for k (positive and nogative) for each value of BsBro. 12) EIGENPUNCTIONS AND RIGENVALUES 87 ‘Thus any solution of the wave equation can be written in a form analogous to Ba, (10.18), ime Wat) = (Jor f ak) Ae ® ule), (12.15) where the Ay are independent of x and ¢; the entire time dependence is, contained in the exponential factor. Equation (12.16) is readily verified to bea solution of Ea. (12.14) by direct substitution, provided that h-S (12.18) ‘The problem of finding the motion of a wave packet is thus resolved into finding the expansion coefficient Ay at some particular time, say t= 0, and then using Eqs. (12.15) and (12.16) to find y(2i) at other times. At ¢ = 0, the exponential factors in Iq (12.15) are unity, and we may use the one-dimensional analogue of the second of Eqs. (11.17) to find As, Ay = Jas@)v(e,0)ee (12.17) ‘The limits on the integral are z= +41, or z= +, necording as box or function normalization is used. ‘The momentum probability fune- a tion P(k) = [Ase * |? = [Ault is independent of the time, 4o that (p) and 4p, for example, are constants. Change with Time of a Minimum Packet. As a simple specific exam- ple, we take Y(z,0) to have tho form of Fa. (12.11) with (2) = (p) = 0, 0 that the wave pucket initially is centered at z = Oand has zero average momentura, ‘Then using box normalization, Hq. (12.17) gives Ag = [2r1%(a0))-# [ie [- zap ste] ae ~ [See] 2.18) where L is assumed to be so large that the contribu from |2| > 41. cun be neglected. Substitution ‘wave function for general values of £ we) = YAc jon to the integral jo Eq. (12.18) gives the Se te) (1219) where k = 2rn/Z and n takes on all positive and nogative integer valuos and zero, As in Seo. 11, L may be taken arbitrarily large, n can be58 QUANTUM MECHANICS [Cnay. IIT regarded as a continuous variable and the summation replaced by {dn, which in turn is the same as (L/2r)/dk. ‘Thus vet) = [SEP [ex [ eae - BE sia ae ‘The position probability density is then welt = [2e[ are + |} sexp, 4 kT (12.81) { 2| (oe) + arta } ‘Equation (12.21) is of the same form as |¥(z,0)|, except that (Az)* isreplaced by (Az)* + M42/4m%(Az)*, which is equal to (Az)* + (Ap)¥/m?, ‘Thus the center of the packet remains at « = 0 while the breadth of the packet increases as ¢ departs from zero in both past and future directions, ‘The emaller the initial uncertainty in position, the larger the uncertainty in momentum and the more rapidly the packet spreads; the time-dopend- ent part of the above expression, (Ap) /m, is simply the distance traveted by a classical particle of momentum Ap in the time t, ‘Use of the é-function normalization does not alter the results of the foregoing calculation, ‘The expression for Ay given in Iq. (12.18) is to be multiplied by (L/2z); in iq, (12,19) the summation is to be replaced directly by fak, thus eliminating a factor 1,/2r; finally, we in Hq. (12.19) is to be multiplied by (L/2r)!. ‘These three factors euncel, and so Eas. (12.20) and (12.21) are unaffected by the choieo of normalization of the ‘momentum eigenfunctions, Classical Limit. We have seen in Sec. 7 that a wave packet always ‘moves like a classical particle in so far us the expectation valuos of its position and momentum are concerned. However, classical dynamics is nly useful as a description of the motion if the spreading of the wave packet can be neglected over times of interest in the particular problem. As asimple example of the kind of parameter that indicates when the classical limit is realized, we consider a wave pucket that corresponds to a classical particle moving in a circular orbit of radius a.nd period 7. We shall assume that this packet is suffctently well localized so that the potential energy does not vary appreciably over its dimensions. ‘Then the classical theory can provide useful description of the motion only if a wave packet like that discussed above spreads by an amount that is.